You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Usage: interlinked [options] [command]
Interlinked CLI: local hooks, activity capture, and developer observability
Options:
-V, --version output the version number
-h, --help display help for command
Commands:
activity [options] Recent activity feed
adopt [options] Bootstrap all ratchet baselines + trigram index from the repo's current state
allowlist Manage the supply-chain package allowlist (.interlinked/package-allowlist.json)
attach [options] Attach local CLI settings to workspace/agent and link remote identity
audit Verify tamper-evidence of the guard-decision audit chain in activity.jsonl
caps [options] View, set, and explain quality caps (lines/function-tokens/complexity/CRAP/coverage)
check [options] Scan project for structural issues and optionally run external tool checks (tsc, biome, eslint, semgrep, gitleaks, mypy, ruff, etc.)
checkpoint [options] [message] Git checkpoint management
ci-status [options] Surface CI failure-rate patterns from GitHub Actions (uses gh CLI)
clean [options] Remove stale data
cloud Inspect the cloud governor (reads cloud_governor from config.local.json)
collect [options] Sync external model sessions (Codex) into .interlinked/timeline.jsonl
compact [options] Gzip + archive the synced prefix of activity.jsonl plus the collection/timeline logs (lossless), reclaiming disk
completions <shell> Output shell completion script (bash, zsh, fish)
context [options] Show effective configuration (merged from all sources)
coverage Per-file coverage ratchet — fails on any file whose coverage drops
cowork Experimental Cowork plugin, native hook evidence, and workspace verification (JSON output)
daemons [options] List active harness daemons, PID liveness, socket paths, and health
data Discover, assess, index and search all local JSONL evidence
deadcode [options] Scan the whole repo for dead-code candidates: unreachable files, unused import bindings, unused exports
debt Inspect pair-scoped TDD debts (coverage / red_suite) from the obligation ledger
design [options] [path] Run Impeccable's deterministic design-slop detector (overused fonts, accent stripes, gradient text, AI palettes, bounce easing, broken images, copy tells) on frontend files. Requires the optional `impeccable` CLI on PATH; degrades gracefully when absent. The built-in `design_slop` check covers a regex subset natively.
disable [options] Stand the harness down for this project (recorded); --uninstall to remove hooks + config
doctest [options] Run doctest-tagged (```bash doctest) code blocks in docs and verify they exit 0
doctor [options] Diagnose issues (local + server checks)
e2e Interlinked's OWN hook/daemon boundary tests (self-test lane); host projects use `interlinked tests e2e`
enable [options] Install hooks + create .interlinked/ config
env [options] Show supported environment variables and their current values
experience Agent-readable trajectory export + analysis (trajectory-v1 interop / trajectory-ix.v1 annotated)
explain [options] Reconstruct what happened (narrative view)
findings Ingest external review findings and drive them to closure
git Git bridge: metadata, trailers, and notes
guard File reservation enforcement via git hooks
handoff [options] <from-agent> <to-agent> Explicit agent-to-agent handoff with context transfer
harness Local harness server: guard evaluation, auto-reservations, agent lifecycle
help [command] display help for command
impact [options] Report evidence-classed local impact facts without causal attribution
inbox [options] Show recent messages from the server
index Manage the trigram search index for grep acceleration
init [options] One-command onboarding: detect clients, configure, login, verify
install-hooks [options] Install agent hooks for detected runners (adapter-based, manifest-driven)
lint Discover and adopt existing lint configurations as Interlinked checks and debt gates
login [options] Authenticate with the server (opens browser)
logout [options] Clear authentication credentials (preserves other config)
logs [options] View local activity log (offline, no server needed)
mcp Record mediated MCP server protocol traffic
metrics [options] Scan the whole codebase: function tokens, companion-test presence, coverage, complexity, and CRAP
mode [options] [name] Show current enforcement mode, or switch to balanced / strict / lenient
multi-edit [options] [path] Apply N old/new string edits atomically to one or more files. Gate runs once on final content. Ambiguity evaluated after prior edits.
mutation Mutation testing: report-score ratchet, local runner measurement, and experimental durable cloud jobs
network Diagnose command-network access and prepare host permission profiles
plan Show agent-emitted plans captured from TaskCreate / ExitPlanMode / structured prompts
query [options] [source] Bounded query over .interlinked JSONL logs (blocks/checks/recurrences/costs/… or a .jsonl path)
recurrence Surface repeating agent behaviors (harness_caught / harness_missed / codebase_existing / tool_failure)
reload [options] One-command dogfood loop: rebuild the CLI from its source checkout, refresh this repo's hooks, restart the daemon — reporting only what actually changed
reminder File reminder management (warnings when files are touched)
replay Capture + replay substrate for the RL/eval environment (G1 inference proxy)
reset [options] Nuclear: clear all local state
resume [options] [checkpoint-id] Resume from latest or specified checkpoint with context
rewind [options] [checkpoint-id] Restore working tree to a checkpoint state
scanner PII filter (content scanner) — toggle, inspect, audit
scratch Manage the sanctioned session/agent-script home (<repo>/scratch/)
search [options] <query> Search the local codebase (ripgrep with native fallback)
semantic Build and query the optional local semantic function index
send [options] <to> [message] Send a message to an agent
setup [options] One-command setup: install hooks, configure server, authenticate
simplify Simplification evidence, explicit recording, and Agent CI handoff preparation
skill Skill marker management (scopes distilled rules via active_when)
spec Spec-substrate utilities (fact ledger, review agenda)
sponsor Opt-in sponsor slot on the statusline (free-sponsor phase)
status [options] Dashboard: local sessions, recent activity, sync status
structure Generic artifact structure management (manifests, catalogs, adoption)
sync [options] Push locally-buffered events to the server
tasks Task management via the server
tdd Inspect and clear TDD cycle state (what the commit gate believes)
telemetry [options] View or tail the local telemetry spool (.interlinked/offline-spool.jsonl)
tests Explain and run affected tests with bounded scheduling and snapshot validation
trace Agent trace export/import
trajectory Inspect trajectory snapshots; replay recorded event streams through sequence detectors
uninstall-hooks [options] Remove hooks previously installed via install-hooks (manifest-driven)
update|upgrade [options] Clone or pull from GitHub, rebuild, and link the CLI
verify [options] [target] Run the default high-signal external catalog plus diff-safe inline checks. Target can be a local path, GitHub URL, or any git remote URL.
verify-changeset [options] Preview the content-quality gate (pre_block + biome + tsc diff-overlay) over a PROPOSED changeset WITHOUT writing — the agent-callable self-gate. Input JSON {version:1, changes:[{path,content}|{path,old_string,new_string}|{path,edits}]} via --file or --stdin.
version Show Interlinked CLI + server version
viz Baseline-test visualizer — the cells, interlinked (loopback dashboard)
watch [options] Monitor server for pending work (messages, tasks, agents)
workspace Registry workspace management (ws_ IDs)
write [options] [path] Write file(s) through the content-quality gate (pre_block + biome + tsc diff-overlay). Supports --stdin, --from-file, and --batch <manifest.json> with rollback protection.
Network
Usage: interlinked network [options] [command]
Diagnose command-network access and prepare host permission profiles
Options:
-h, --help display help for command
Commands:
check [options] <endpoints...> Probe HTTP(S) origins from this process,
without credentials or inference
codex-profile [options] <name> Preview a reusable workspace-restricted Codex
network profile
help [command] display help for command
network check
Usage: interlinked network check [options] <endpoints...>
Probe HTTP(S) origins from this process, without credentials or inference
Options:
--json Machine-readable runtime evidence
-h, --help display help for command
network codex-profile
Usage: interlinked network codex-profile [options] <name>
Preview a reusable workspace-restricted Codex network profile
Options:
--allow <endpoints...> HTTP(S) endpoints whose exact hosts may be reached
--write Create $CODEX_HOME/<name>.config.toml; refuse
conflicting existing files
--json Machine-readable output
-h, --help display help for command
Data
Usage: interlinked data [options] [command]
Discover, assess, index and search all local JSONL evidence
Options:
-h, --help display help for command
Commands:
audit Locate historical audit failures and record
explicit observation boundaries
maintain [options] Plan evidence-class retention; optionally
index and losslessly rotate supported logs
configure [options] Show or update local data automation and
import budgets
lab Explicit, isolated evidence storage
experiments; preserves original logs and
indexes
scan [options] [text] Search bounded live JSONL/gzip directly,
without SQLite or copying logs
recurrence-inventory [options] Latest scoped finding inventory, distinct
from incident counts
investigate [options] Correlate session/call/file evidence, missing
phases and folded file obligations
catalog [options] Registered source contracts and recursively
discovered files
health [options] Producer receipts, unknowns, storage and
index coverage
status [options] Index freshness, incomplete sources and parse
failures
index [options] Incrementally index live files and gzip
archives; repeat to resume
search [options] [text] Full-text and structured search over the
local index
show [options] <id> Retrieve original evidence bytes and verify
the indexed hash
sessions [options] Indexed sessions evidence with coverage and
provenance
files [options] Indexed files evidence with coverage and
provenance
checks [options] Indexed checks evidence with coverage and
provenance
usage [options] Indexed usage evidence with coverage and
provenance
schema [options] Indexed schema evidence with coverage and
provenance
suggestions [options] Indexed suggestions evidence with coverage
and provenance
help [command] display help for command
data catalog
Usage: interlinked data catalog [options]
Registered source contracts and recursively discovered files
Options:
--cwd <path> Project root
--json Machine-readable output
--short Compact JSON
-h, --help display help for command
data health
Usage: interlinked data health [options]
Producer receipts, unknowns, storage and index coverage
Options:
--cwd <path> Project root
--json Machine-readable output
--short Compact JSON
-h, --help display help for command
data status
Usage: interlinked data status [options]
Index freshness, incomplete sources and parse failures
Options:
--cwd <path> Project root
--json Machine-readable output
--short Compact JSON
-h, --help display help for command
data index
Usage: interlinked data index [options]
Incrementally index live files and gzip archives; repeat to resume
Options:
--max-mb <n> Expanded byte budget per run (configured default: 256)
--max-records <n> Record budget per run (configured default: 250000)
--rebuild Clear and rebuild only the derived index; raw evidence is
preserved
--source <name> One logical source
--no-archives Skip archive segments
--cwd <path> Project root
--json Machine-readable output
--short Compact JSON
-h, --help display help for command
data scan
Usage: interlinked data scan [options] [text]
Search bounded live JSONL/gzip directly, without SQLite or copying logs
Options:
--source <value> Exact source filter
--category <value> Exact category filter
--session <value> Exact session filter
--actor <value> Exact actor filter
--provider <value> Exact provider filter
--model <value> Exact model filter
--file <value> Exact file filter
--check <value> Exact check filter
--kind <value> Exact kind filter
--decision <value> Exact decision filter
--origin <value> Exact origin filter
--call <value> Exact call filter
--since <time> Event time: duration or ISO timestamp; excludes undated
records
--until <time> Latest event time: duration or ISO timestamp
--limit <n> Maximum results (1..1000)
--no-archives Only retained live-source evidence
--cwd <path> Project root
--json Machine-readable output
--short Compact JSON
--max-mb <n> Expanded scan budget (default 32, maximum 1024 MiB)
--max-records <n> Physical line budget (default 25000)
--offset <n> Result offset within the scanned scope
--raw Include original record text and hash in returned rows
--full-text Match all decoded string values, beyond the bounded index
projection
-h, --help display help for command
data lab
Usage: interlinked data lab [options] [command]
Explicit, isolated evidence storage experiments; preserves original logs and
indexes
Options:
-h, --help display help for command
Commands:
generate [options] Evidence experiment: generate
snapshot [options] Evidence experiment: snapshot
build [options] Evidence experiment: build
search [options] Evidence experiment: search
show [options] Evidence experiment: show
fts [options] Evidence experiment: fts
analytics-export [options] Evidence experiment: analytics-export
analytics-query [options] Evidence experiment: analytics-query
benchmark [options] Evidence experiment: benchmark
worker [options] Evidence experiment: worker
help [command] display help for command
data search
Usage: interlinked data search [options] [text]
Full-text and structured search over the local index
Options:
--source <value> Exact source filter
--category <value> Exact category filter
--session <value> Exact session filter
--actor <value> Exact actor filter
--provider <value> Exact provider filter
--model <value> Exact model filter
--file <value> Exact file filter
--check <value> Exact check filter
--kind <value> Exact kind filter
--decision <value> Exact decision filter
--origin <value> Exact origin filter
--call <value> Exact call filter
--since <time> Event time: duration or ISO timestamp; excludes undated
records
--until <time> Latest event time: duration or ISO timestamp
--limit <n> Maximum results (1..1000)
--no-archives Only retained live-source evidence
--cwd <path> Project root
--json Machine-readable output
--short Compact JSON
--fts Interpret text as SQLite FTS5 grammar
--offset <n> Result offset
-h, --help display help for command
data show
Usage: interlinked data show [options] <id>
Retrieve original evidence bytes and verify the indexed hash
Options:
--cwd <path> Project root
--json Machine-readable output
--short Compact JSON
-h, --help display help for command
data investigate
Usage: interlinked data investigate [options]
Correlate session/call/file evidence, missing phases and folded file
obligations
Options:
--source <value> Exact source filter
--category <value> Exact category filter
--session <value> Exact session filter
--actor <value> Exact actor filter
--provider <value> Exact provider filter
--model <value> Exact model filter
--file <value> Exact file filter
--check <value> Exact check filter
--kind <value> Exact kind filter
--decision <value> Exact decision filter
--origin <value> Exact origin filter
--call <value> Exact call filter
--since <time> Event time: duration or ISO timestamp; excludes undated
records
--until <time> Latest event time: duration or ISO timestamp
--limit <n> Maximum results (1..1000)
--no-archives Only retained live-source evidence
--cwd <path> Project root
--json Machine-readable output
--short Compact JSON
-h, --help display help for command
data sessions
Usage: interlinked data sessions [options]
Indexed sessions evidence with coverage and provenance
Options:
--source <value> Exact source filter
--category <value> Exact category filter
--session <value> Exact session filter
--actor <value> Exact actor filter
--provider <value> Exact provider filter
--model <value> Exact model filter
--file <value> Exact file filter
--check <value> Exact check filter
--kind <value> Exact kind filter
--decision <value> Exact decision filter
--origin <value> Exact origin filter
--call <value> Exact call filter
--since <time> Event time: duration or ISO timestamp; excludes undated
records
--until <time> Latest event time: duration or ISO timestamp
--limit <n> Maximum results (1..1000)
--no-archives Only retained live-source evidence
--cwd <path> Project root
--json Machine-readable output
--short Compact JSON
-h, --help display help for command
data files
Usage: interlinked data files [options]
Indexed files evidence with coverage and provenance
Options:
--source <value> Exact source filter
--category <value> Exact category filter
--session <value> Exact session filter
--actor <value> Exact actor filter
--provider <value> Exact provider filter
--model <value> Exact model filter
--file <value> Exact file filter
--check <value> Exact check filter
--kind <value> Exact kind filter
--decision <value> Exact decision filter
--origin <value> Exact origin filter
--call <value> Exact call filter
--since <time> Event time: duration or ISO timestamp; excludes undated
records
--until <time> Latest event time: duration or ISO timestamp
--limit <n> Maximum results (1..1000)
--no-archives Only retained live-source evidence
--cwd <path> Project root
--json Machine-readable output
--short Compact JSON
-h, --help display help for command
data checks
Usage: interlinked data checks [options]
Indexed checks evidence with coverage and provenance
Options:
--source <value> Exact source filter
--category <value> Exact category filter
--session <value> Exact session filter
--actor <value> Exact actor filter
--provider <value> Exact provider filter
--model <value> Exact model filter
--file <value> Exact file filter
--check <value> Exact check filter
--kind <value> Exact kind filter
--decision <value> Exact decision filter
--origin <value> Exact origin filter
--call <value> Exact call filter
--since <time> Event time: duration or ISO timestamp; excludes undated
records
--until <time> Latest event time: duration or ISO timestamp
--limit <n> Maximum results (1..1000)
--no-archives Only retained live-source evidence
--cwd <path> Project root
--json Machine-readable output
--short Compact JSON
-h, --help display help for command
data usage
Usage: interlinked data usage [options]
Indexed usage evidence with coverage and provenance
Options:
--source <value> Exact source filter
--category <value> Exact category filter
--session <value> Exact session filter
--actor <value> Exact actor filter
--provider <value> Exact provider filter
--model <value> Exact model filter
--file <value> Exact file filter
--check <value> Exact check filter
--kind <value> Exact kind filter
--decision <value> Exact decision filter
--origin <value> Exact origin filter
--call <value> Exact call filter
--since <time> Event time: duration or ISO timestamp; excludes undated
records
--until <time> Latest event time: duration or ISO timestamp
--limit <n> Maximum results (1..1000)
--no-archives Only retained live-source evidence
--cwd <path> Project root
--json Machine-readable output
--short Compact JSON
-h, --help display help for command
data schema
Usage: interlinked data schema [options]
Indexed schema evidence with coverage and provenance
Options:
--source <value> Exact source filter
--category <value> Exact category filter
--session <value> Exact session filter
--actor <value> Exact actor filter
--provider <value> Exact provider filter
--model <value> Exact model filter
--file <value> Exact file filter
--check <value> Exact check filter
--kind <value> Exact kind filter
--decision <value> Exact decision filter
--origin <value> Exact origin filter
--call <value> Exact call filter
--since <time> Event time: duration or ISO timestamp; excludes undated
records
--until <time> Latest event time: duration or ISO timestamp
--limit <n> Maximum results (1..1000)
--no-archives Only retained live-source evidence
--cwd <path> Project root
--json Machine-readable output
--short Compact JSON
-h, --help display help for command
data suggestions
Usage: interlinked data suggestions [options]
Indexed suggestions evidence with coverage and provenance
Options:
--source <value> Exact source filter
--category <value> Exact category filter
--session <value> Exact session filter
--actor <value> Exact actor filter
--provider <value> Exact provider filter
--model <value> Exact model filter
--file <value> Exact file filter
--check <value> Exact check filter
--kind <value> Exact kind filter
--decision <value> Exact decision filter
--origin <value> Exact origin filter
--call <value> Exact call filter
--since <time> Event time: duration or ISO timestamp; excludes undated
records
--until <time> Latest event time: duration or ISO timestamp
--limit <n> Maximum results (1..1000)
--no-archives Only retained live-source evidence
--cwd <path> Project root
--json Machine-readable output
--short Compact JSON
-h, --help display help for command
data recurrence-inventory
Usage: interlinked data recurrence-inventory [options]
Latest scoped finding inventory, distinct from incident counts
Options:
--cwd <path> Project root
--json Machine-readable output
--short Compact JSON
-h, --help display help for command
data configure
Usage: interlinked data configure [options]
Show or update local data automation and import budgets
Options:
--auto-index <on|off> Index in the governed background lane at session end
--auto-compact <on|off> Losslessly rotate supported event logs during
maintenance
--index-mb <n> Expanded byte budget per index pass
--index-records <n> Record budget per index pass
--keep-live-mb <n> Retained live tail size
--compact-at-mb <n> Live-size threshold for eligible logs
--cwd <path> Project root
--json Machine-readable output
-h, --help display help for command
data maintain
Usage: interlinked data maintain [options]
Plan evidence-class retention; optionally index and losslessly rotate supported
logs
Options:
--execute Run explicitly requested or configured maintenance
--index Also run the SQLite importer (independent of rotation)
--no-index Skip the importer even if automatic indexing is configured
--compact Also rotate eligible collection/timeline logs; retains gzip
evidence
--cwd <path> Project root
--json Machine-readable output
-h, --help display help for command
data audit
Usage: interlinked data audit [options] [command]
Locate historical audit failures and record explicit observation boundaries
Options:
-h, --help display help for command
Commands:
verify [options] Verify retained evidence after an explicit checkpoint,
including after rotation
diagnose [options] Verify archives/live data and locate the first failing
physical record
checkpoint [options] Record a verified live-tail boundary; historical
verdict stays unchanged
help [command] display help for command
Structure
Usage: interlinked structure [options] [command]
Generic artifact structure management (manifests, catalogs, adoption)
Options:
-h, --help display help for command
Commands:
init [options] Create interlinked/structure.json and scaffold
artifact files
scan [options] Build or refresh local generated artifact
catalogs
status [options] Show adoption coverage, cache staleness, and
invalid references
accept [options] Promote extracted findings into committed
artifact files
doctor [options] Validate structure files, cache freshness, and
cross-references
baseline [options] <action> Manage structure baselines (save, clear, status)
help [command] display help for command
structure init
Usage: interlinked structure init [options]
Create interlinked/structure.json and scaffold artifact files
Options:
--mode <mode> Structure mode: minimal, standard, strict (default:
"standard")
--with <categories> Comma-separated artifact categories to scaffold
--write Actually write files (default is dry-run)
--json Machine-readable output
-h, --help display help for command
structure scan
Usage: interlinked structure scan [options]
Build or refresh local generated artifact catalogs
Options:
--full Force full rescan
--incremental Only refresh changed categories
--json Machine-readable output
-h, --help display help for command
structure status
Usage: interlinked structure status [options]
Show adoption coverage, cache staleness, and invalid references
Options:
--json Machine-readable output
-h, --help display help for command
structure accept
Usage: interlinked structure accept [options]
Promote extracted findings into committed artifact files
Options:
--json Machine-readable output
-h, --help display help for command
structure doctor
Usage: interlinked structure doctor [options]
Validate structure files, cache freshness, and cross-references
Options:
--json Machine-readable output
-h, --help display help for command
Usage: interlinked verify [options] [target]
Run the default high-signal external catalog plus diff-safe inline checks.
Target can be a local path, GitHub URL, or any git remote URL.
Options:
--only <tool> Run one named external check (for example: tsc,
biome, oxlint, eslint, semgrep, gitleaks, knip,
docs-check, sca)
--suggestions Also run scored regex heuristics (sql-injection,
perf, quality)
--json Machine-readable output
--details Show per-file details for all findings
--cwd <path> Project root (default: current directory)
--branch <ref> Branch, tag, or commit to check (remote repos)
--subdir <path> Only scan a subdirectory (useful for monorepos)
--skip <checks> Skip specific checks (comma-separated:
semgrep,knip,complexity,silent_catches,...)
--suppress <entries...> Suppress a finding: file:check or file:check:reason
--show-suppressions List all active suppressions
--structure Include generic artifact structure checks
--structure-only Run only structure checks
--adoption-gate Fail when adopted categories drop below thresholds
--all-checks Include broad advisory smell checks and dead-code
scans in addition to the default high-signal audit
--dead-code Run Supermodel's cloud dead-code analysis (opt-in;
requires the `supermodel` CLI)
--fp-filter Report per check how many [heuristic] findings the
E18 fp_filter would suppress (also on when
fp_filter.enabled)
-h, --help display help for command
Lint
Usage: interlinked lint [options] [command]
Discover and adopt existing lint configurations as Interlinked checks and debt
gates
Options:
-h, --help display help for command
Commands:
scan [options] [target] Inventory lint configs, declarations, ignores and
scripts across nested packages
import [options] [target] Preview or apply lint adoption while preserving
original analyzer semantics
check [options] [target] Gate imported lint findings: exit 1 for new debt,
2 when no complete verdict exists
help [command] display help for command
lint scan
Usage: interlinked lint scan [options] [target]
Inventory lint configs, declarations, ignores and scripts across nested
packages
Options:
--json Machine-readable inventory and import plan
--details Show bounded source declaration excerpts
-h, --help display help for command
lint import
Usage: interlinked lint import [options] [target]
Preview or apply lint adoption while preserving original analyzer semantics
Options:
--write Write the import policy and enable its
PostToolUse check
--baseline With --write, run analyzers and seed or tighten
existing-debt allowances
--eslint-config <file> Select an ESLint config relative to target
(repeatable) (default: [])
--eslint-scope <directory> Working directory and lint target for selected
ESLint configs, relative to target (default: .)
--config <tool=file> Select any supported analyzer config (repeatable)
(default: [])
--scope <directory> Working directory for --config selections
(default: .)
--target <path> Analyzer target relative to --scope for explicit
--config selections (repeatable) (default: [])
--only-selected Add only explicit configs, retaining existing
profiles; cadence applies only to selections
--cadence <cadence> Set imported profiles to hook or audit cadence
--gate <policy> Gate all findings (default), or errors; native
warnings stay advisory with errors
--timeout <ms> Total analyzer batch budget, up to 300000 ms
(default: "30000")
--json Machine-readable import plan/result
--details Show bounded source declaration excerpts
-h, --help display help for command
lint check
Usage: interlinked lint check [options] [target]
Gate imported lint findings: exit 1 for new debt, 2 when no complete verdict
exists
Options:
--update-baseline Seed unadopted scopes and tighten existing allowances
after complete analysis
--timeout <ms> Total analyzer batch budget, up to 300000 ms (default:
"30000")
--json Machine-readable findings and measurement status
-h, --help display help for command
Metrics
Usage: interlinked metrics [options] [command]
Scan the whole codebase: function tokens, companion-test presence, coverage,
complexity, and CRAP
Options:
--cwd <path> Project root (default: current directory)
--top <n> Number of function/file token and CRAP
hotspots to show (default: 25)
--include-tests Include test/spec functions as advisory
token measurements
--json Machine-readable output (full per-file +
per-function)
--short One-line summary
--full Show every per-file and per-function
token measurement
-h, --help display help for command
Commands:
coupling [options] Change coupling from git history —
co-changed file pairs; pairs with no
import edge are flagged 'hidden'
arch [options] Martin metrics per directory
(Ca/Ce/instability) + propagation cost
from the import graph
rework [options] Churn age from git blame — share of
changed lines whose previous version was
written in the last --window days
stale-readme [options] READMEs the code moved past — commits
touching each README's directory since
the README last changed (telemetry, never
blocks)
complexity [options] Complexity census: percentiles,
histograms, top-N hotspots, per-file
mass, and over-cap counts for cyclomatic
/ cognitive / lines
split-plan [options] <file> Where to cut one over-cap file:
intra-file reference graph (TS AST) → 2–4
cohesive modules with line count, ΣCC,
imports, a suggested filename each, and
the cross-module references the split
creates
score [options] Explained 0–100 quality burden with
explicit evidence completeness; no model
calls
jit [options] [ref] Just-in-time commit risk (Kamei 2013):
score one commit or the staged diff and
rank it against the repo's own recent
commits
expressions [options] [target] Measure JS/TS expression tokens,
callbacks and presentation; advisory
findings
diagnostics [options] Explained verbosity and erosion census;
advisory evidence, no quality verdict
gates [options] Show disabled gates, actual coverage
executions, freshness and runtime
percentiles
coverage Operate the exact Vitest contribution
index used by per-edit coverage
corpus [options] <manifest> Score pinned clean local repositories and
persist reproducible reports
catalog [options] List metric contracts and every
check/guard disposition
explain [options] <metric> Explain one metric's denominator, burden
and composite weight
compare [options] <before> <after> Compare saved score JSON with profile and
evidence compatibility checks
deletions [options] Join redundancy, reachability, coverage
and survivor review candidates
evidence Bound coverage and mutation evidence to
source, tests, runner and artifact hashes
decisions Experimental semantic measurements with
local decision models, ThinkThen, and
auditable uncertainty
metrics coupling
Usage: interlinked metrics coupling [options]
Change coupling from git history — co-changed file pairs; pairs with no import
edge are flagged 'hidden'
Options:
--cwd <path> Project root (default: current directory)
--since <when> git --since expression (default: '90 days ago')
--min-support <n> Minimum co-change commits per pair (default: 4)
--max-commit-files <n> Skip bulk commits touching more files (default: 30)
--min-strength <pct> Minimum Tornhill strength percentage (default: 30)
--limit <n> Maximum pairs to report (default: 25)
--json Machine-readable output
--short One-line summary
-h, --help display help for command
metrics arch
Usage: interlinked metrics arch [options]
Martin metrics per directory (Ca/Ce/instability) + propagation cost from the
import graph
Options:
--cwd <path> Project root (default: current directory)
--depth <n> Directory fold depth (default: 2)
--include-tests Include test files in the edge set
--json Machine-readable output
--short One-line summary
-h, --help display help for command
metrics rework
Usage: interlinked metrics rework [options]
Churn age from git blame — share of changed lines whose previous version was
written in the last --window days
Options:
--cwd <path> Project root (default: current directory)
--days <n> How far back to scan commits (default: 30)
--window <n> Rework age threshold in days (default: 14)
--max-commits <n> Commit scan cap (default: 100)
--max-commit-files <n> Skip bulk commits touching more files (default: 30)
--json Machine-readable output
--short One-line summary
-h, --help display help for command
metrics complexity
Usage: interlinked metrics complexity [options]
Complexity census: percentiles, histograms, top-N hotspots, per-file mass, and
over-cap counts for cyclomatic / cognitive / lines
Options:
--cwd <path> Project root (default: current directory)
--top <n> Hotspots per metric and files by mass (default: 20)
--metric <name> cyclomatic | cognitive | lines | all (default: all)
--churn-days <n> Hotspot window: ΣCC × lines changed in the last n days
(default: 90)
--json Machine-readable output
--short One-line summary
-h, --help display help for command
metrics split-plan
Usage: interlinked metrics split-plan [options] <file>
Where to cut one over-cap file: intra-file reference graph (TS AST) → 2–4
cohesive modules with line count, ΣCC, imports, a suggested filename each, and
the cross-module references the split creates
Options:
--cwd <path> Project root (default: current directory)
--max-clusters <n> Upper bound on proposed modules, 2–4 (default: 4)
--json Machine-readable output
--short One-line summary
-h, --help display help for command
metrics score
Usage: interlinked metrics score [options]
Explained 0–100 quality burden with explicit evidence completeness; no model
calls
Options:
--profile <name> slop-v1 (default) or legacy structure-v1
--cwd <path> Project root (default: current directory)
--json Full measurements, profile, hashes and explicit evidence
gaps
--short One-line score and evidence status
-h, --help display help for command
metrics diagnostics
Usage: interlinked metrics diagnostics [options] [command]
Explained verbosity and erosion census; advisory evidence, no quality verdict
Options:
--cwd <path> Repository root
--json Machine-readable result
--short Compact summary
--profile <name> js-ts (default) or python; separate
counting contracts
-h, --help display help for command
Commands:
compare [options] <before> <after> Compare saved diagnostic reports with
scope and identity checks
metrics diagnostics compare
Usage: interlinked metrics diagnostics compare [options] <before> <after>
Compare saved diagnostic reports with scope and identity checks
Options:
--cwd <path> Repository root
--json Machine-readable result
--short Compact summary
-h, --help display help for command
metrics catalog
Usage: interlinked metrics catalog [options]
List metric contracts and every check/guard disposition
Options:
--cwd <path> Repository root
--json Machine-readable result
--short Compact summary
--checks Include individual check and guard dispositions
-h, --help display help for command
metrics explain
Usage: interlinked metrics explain [options] <metric>
Explain one metric's denominator, burden and composite weight
Options:
--cwd <path> Repository root
--json Machine-readable result
--short Compact summary
-h, --help display help for command
metrics compare
Usage: interlinked metrics compare [options] <before> <after>
Compare saved score JSON with profile and evidence compatibility checks
Options:
--cwd <path> Repository root
--json Machine-readable result
--short Compact summary
-h, --help display help for command
metrics corpus
Usage: interlinked metrics corpus [options] <manifest>
Score pinned clean local repositories and persist reproducible reports
Options:
--cwd <path> Repository root
--json Machine-readable result
--short Compact summary
--out <directory> Directory for per-repository reports and corpus summary
-h, --help display help for command
metrics gates
Usage: interlinked metrics gates [options]
Show disabled gates, actual coverage executions, freshness and runtime
percentiles
Options:
--cwd <path> Repository root
--json Machine-readable result
--short Compact summary
-h, --help display help for command
metrics deletions
Usage: interlinked metrics deletions [options] [command]
Join redundancy, reachability, coverage and survivor review candidates
Options:
--cwd <path> Repository root
--json Machine-readable result
--short Compact summary
-h, --help display help for command
Commands:
validate [options] <plan> Test an explicit removal plan in an isolated copy;
preserve the repository
metrics deletions validate
Usage: interlinked metrics deletions validate [options] <plan>
Test an explicit removal plan in an isolated copy; preserve the repository
Options:
--cwd <path> Repository root
--json Machine-readable result
--short Compact summary
--timeout <ms> Total trial budget (default: "60000")
-h, --help display help for command
metrics evidence
Usage: interlinked metrics evidence [options] [command]
Bound coverage and mutation evidence to source, tests, runner and artifact
hashes
Options:
-h, --help display help for command
Commands:
status [options] List current, stale and inconclusive behavioral receipts
identity [options] Export current input hashes for CI receipts
import [options] <receipt> <artifact> Validate and import a CI-produced receipt and report
run [options] Run an explicit command in a bounded isolated copy; no model calls by Interlinked
help [command] display help for command
metrics evidence status
Usage: interlinked metrics evidence status [options]
List current, stale and inconclusive behavioral receipts
Options:
--cwd <path> Repository root
--json Machine-readable result
--short Compact summary
-h, --help display help for command
metrics evidence identity
Usage: interlinked metrics evidence identity [options]
Export current input hashes for CI receipts
Options:
--cwd <path> Repository root
--json Machine-readable result
--short Compact summary
-h, --help display help for command
metrics evidence import
Usage: interlinked metrics evidence import [options] <receipt> <artifact>
Validate and import a CI-produced receipt and report
Options:
--cwd <path> Repository root
--json Machine-readable result
--short Compact summary
-h, --help display help for command
metrics evidence run
Usage: interlinked metrics evidence run [options]
Run an explicit command in a bounded isolated copy; no model calls by
Interlinked
Options:
--cwd <path> Repository root
--json Machine-readable result
--short Compact summary
--kind <kind> coverage or mutation
--command <json> JSON argv array, executed without a shell
--artifact <path> Repository-relative Istanbul or mutation JSON
output
--runner-version <version> Declared runner version
--policy <id> Coverage/mutation operator policy identity
--timeout <ms> Total copy and execution budget (default:
"60000")
--resume Reuse a passing receipt only when all inputs and
runner identity match
-h, --help display help for command
metrics coverage
Usage: interlinked metrics coverage [options] [command]
Operate the exact Vitest contribution index used by per-edit coverage
Options:
-h, --help display help for command
Commands:
warm [options] Measure the full suite in an overlay and initialize the
contribution index
status [options] Validate coverage index inputs and report stale shards
help [command] display help for command
metrics coverage status
Usage: interlinked metrics coverage status [options]
Validate coverage index inputs and report stale shards
Options:
--cwd <path> Repository root
--json Machine-readable result
--short Compact summary
--timeout <ms> Coverage execution budget (default: "60000")
-h, --help display help for command
metrics coverage warm
Usage: interlinked metrics coverage warm [options]
Measure the full suite in an overlay and initialize the contribution index
Options:
--cwd <path> Repository root
--json Machine-readable result
--short Compact summary
--timeout <ms> Coverage execution budget (default: "60000")
-h, --help display help for command
metrics decisions
Usage: interlinked metrics decisions [options] [command]
Experimental semantic measurements with local decision models, ThinkThen, and
auditable uncertainty
Options:
-h, --help display help for command
Commands:
init [options] Write a disabled, local-first decision
configuration
catalog [options] Show versioned metric contracts and the
ThinkThen question set
prepare [options] Extract touched function/hunk evidence from
git without checkout
snapshot [options] Index function evidence and import edges at a
pinned Git revision
candidates [options] <input> Build an obligation cohort including consumers
not retrieved
retrieve [options] <input> Attach time-safe BM25, embedding, or hybrid
historical neighbors
pilot [options] Sample real changes and retain unavailable
revision counts
labels [options] <input> Create a blind independent-review label
template from a run
join-labels [options] <input> Join complete adjudicated labels with saved
probabilities
probe [options] <input> Run fresh perturbation, repetition, and
batch-composition probes
measure [options] <input> Measure an evidence population and retain raw
probabilities
review [options] <input> Run relational obligations, atomicity,
faithfulness, or comparison
embed [options] <input> Embed evidence locally, then unload the model
geometry [options] <input> Measure provisional embedding churn,
dispersion, and nearest-neighbor outliers
audit [options] <input> Evaluate a full labeled cohort with a temporal
split and baselines
replay [options] <input> Reinterpret a retained run at a new
uncertainty band, offline
diff [options] <input> Compare retained answers and report
model/question drift
help [command] display help for command
metrics decisions init
Usage: interlinked metrics decisions init [options]
Write a disabled, local-first decision configuration
Options:
--cwd <path> Project root
--config <file> Explicit decision configuration
--output <file> Write an immutable JSON artifact
--metrics <ids> Comma-separated metric contract IDs
--band <low:high> Uncertainty band; probabilities are reusable
--command <path> ThinkThen executable
--max-records <n> Record budget; excluded records remain unassessed
--plan Print native request plans without inference
--fresh Fresh inference; preserve the previous cached
observation
--base <ref> Git base revision
--revision <ref> Committed revision; default is working tree
--file <paths...> Restrict changed source files
--message <text> Description of a working-tree change
--analysis <kind> atomicity, faithfulness, obligations, compare,
coupling, rework
--claims <file> Literal description claims as a JSON array
--cutoff <date> Train before this date; evaluate at/after it
--after <file> Second retained run for diff
--corpus <file> Snapshot or historical function corpus
--retrieval <method> lexical, embedding, or hybrid
--count <n> Pilot commit budget (1..1000)
--before <date> Pilot history upper date bound
--labels <file> Independent adjudicated label file
--max-files <n> Snapshot file budget (1..10000)
--unit <kind> Prepared evidence: function, hunk, or all
--bca-command <path> Optional pinned bca executable for preparation
--bca-version <text> Exact required bca --version stdout
--comparison <file> Matched retrieval-arm labeled cohort for five-arm audit
--json Machine-readable output (always enabled)
-h, --help display help for command
metrics decisions catalog
Usage: interlinked metrics decisions catalog [options]
Show versioned metric contracts and the ThinkThen question set
Options:
--cwd <path> Project root
--config <file> Explicit decision configuration
--output <file> Write an immutable JSON artifact
--metrics <ids> Comma-separated metric contract IDs
--band <low:high> Uncertainty band; probabilities are reusable
--command <path> ThinkThen executable
--max-records <n> Record budget; excluded records remain unassessed
--plan Print native request plans without inference
--fresh Fresh inference; preserve the previous cached
observation
--base <ref> Git base revision
--revision <ref> Committed revision; default is working tree
--file <paths...> Restrict changed source files
--message <text> Description of a working-tree change
--analysis <kind> atomicity, faithfulness, obligations, compare,
coupling, rework
--claims <file> Literal description claims as a JSON array
--cutoff <date> Train before this date; evaluate at/after it
--after <file> Second retained run for diff
--corpus <file> Snapshot or historical function corpus
--retrieval <method> lexical, embedding, or hybrid
--count <n> Pilot commit budget (1..1000)
--before <date> Pilot history upper date bound
--labels <file> Independent adjudicated label file
--max-files <n> Snapshot file budget (1..10000)
--unit <kind> Prepared evidence: function, hunk, or all
--bca-command <path> Optional pinned bca executable for preparation
--bca-version <text> Exact required bca --version stdout
--comparison <file> Matched retrieval-arm labeled cohort for five-arm audit
--json Machine-readable output (always enabled)
-h, --help display help for command
metrics decisions prepare
Usage: interlinked metrics decisions prepare [options]
Extract touched function/hunk evidence from git without checkout
Options:
--cwd <path> Project root
--config <file> Explicit decision configuration
--output <file> Write an immutable JSON artifact
--metrics <ids> Comma-separated metric contract IDs
--band <low:high> Uncertainty band; probabilities are reusable
--command <path> ThinkThen executable
--max-records <n> Record budget; excluded records remain unassessed
--plan Print native request plans without inference
--fresh Fresh inference; preserve the previous cached
observation
--base <ref> Git base revision
--revision <ref> Committed revision; default is working tree
--file <paths...> Restrict changed source files
--message <text> Description of a working-tree change
--analysis <kind> atomicity, faithfulness, obligations, compare,
coupling, rework
--claims <file> Literal description claims as a JSON array
--cutoff <date> Train before this date; evaluate at/after it
--after <file> Second retained run for diff
--corpus <file> Snapshot or historical function corpus
--retrieval <method> lexical, embedding, or hybrid
--count <n> Pilot commit budget (1..1000)
--before <date> Pilot history upper date bound
--labels <file> Independent adjudicated label file
--max-files <n> Snapshot file budget (1..10000)
--unit <kind> Prepared evidence: function, hunk, or all
--bca-command <path> Optional pinned bca executable for preparation
--bca-version <text> Exact required bca --version stdout
--comparison <file> Matched retrieval-arm labeled cohort for five-arm audit
--json Machine-readable output (always enabled)
-h, --help display help for command
metrics decisions snapshot
Usage: interlinked metrics decisions snapshot [options]
Index function evidence and import edges at a pinned Git revision
Options:
--cwd <path> Project root
--config <file> Explicit decision configuration
--output <file> Write an immutable JSON artifact
--metrics <ids> Comma-separated metric contract IDs
--band <low:high> Uncertainty band; probabilities are reusable
--command <path> ThinkThen executable
--max-records <n> Record budget; excluded records remain unassessed
--plan Print native request plans without inference
--fresh Fresh inference; preserve the previous cached
observation
--base <ref> Git base revision
--revision <ref> Committed revision; default is working tree
--file <paths...> Restrict changed source files
--message <text> Description of a working-tree change
--analysis <kind> atomicity, faithfulness, obligations, compare,
coupling, rework
--claims <file> Literal description claims as a JSON array
--cutoff <date> Train before this date; evaluate at/after it
--after <file> Second retained run for diff
--corpus <file> Snapshot or historical function corpus
--retrieval <method> lexical, embedding, or hybrid
--count <n> Pilot commit budget (1..1000)
--before <date> Pilot history upper date bound
--labels <file> Independent adjudicated label file
--max-files <n> Snapshot file budget (1..10000)
--unit <kind> Prepared evidence: function, hunk, or all
--bca-command <path> Optional pinned bca executable for preparation
--bca-version <text> Exact required bca --version stdout
--comparison <file> Matched retrieval-arm labeled cohort for five-arm audit
--json Machine-readable output (always enabled)
-h, --help display help for command
metrics decisions candidates
Usage: interlinked metrics decisions candidates [options] <input>
Build an obligation cohort including consumers not retrieved
Options:
--cwd <path> Project root
--config <file> Explicit decision configuration
--output <file> Write an immutable JSON artifact
--metrics <ids> Comma-separated metric contract IDs
--band <low:high> Uncertainty band; probabilities are reusable
--command <path> ThinkThen executable
--max-records <n> Record budget; excluded records remain unassessed
--plan Print native request plans without inference
--fresh Fresh inference; preserve the previous cached
observation
--base <ref> Git base revision
--revision <ref> Committed revision; default is working tree
--file <paths...> Restrict changed source files
--message <text> Description of a working-tree change
--analysis <kind> atomicity, faithfulness, obligations, compare,
coupling, rework
--claims <file> Literal description claims as a JSON array
--cutoff <date> Train before this date; evaluate at/after it
--after <file> Second retained run for diff
--corpus <file> Snapshot or historical function corpus
--retrieval <method> lexical, embedding, or hybrid
--count <n> Pilot commit budget (1..1000)
--before <date> Pilot history upper date bound
--labels <file> Independent adjudicated label file
--max-files <n> Snapshot file budget (1..10000)
--unit <kind> Prepared evidence: function, hunk, or all
--bca-command <path> Optional pinned bca executable for preparation
--bca-version <text> Exact required bca --version stdout
--comparison <file> Matched retrieval-arm labeled cohort for five-arm audit
--json Machine-readable output (always enabled)
-h, --help display help for command
metrics decisions retrieve
Usage: interlinked metrics decisions retrieve [options] <input>
Attach time-safe BM25, embedding, or hybrid historical neighbors
Options:
--cwd <path> Project root
--config <file> Explicit decision configuration
--output <file> Write an immutable JSON artifact
--metrics <ids> Comma-separated metric contract IDs
--band <low:high> Uncertainty band; probabilities are reusable
--command <path> ThinkThen executable
--max-records <n> Record budget; excluded records remain unassessed
--plan Print native request plans without inference
--fresh Fresh inference; preserve the previous cached
observation
--base <ref> Git base revision
--revision <ref> Committed revision; default is working tree
--file <paths...> Restrict changed source files
--message <text> Description of a working-tree change
--analysis <kind> atomicity, faithfulness, obligations, compare,
coupling, rework
--claims <file> Literal description claims as a JSON array
--cutoff <date> Train before this date; evaluate at/after it
--after <file> Second retained run for diff
--corpus <file> Snapshot or historical function corpus
--retrieval <method> lexical, embedding, or hybrid
--count <n> Pilot commit budget (1..1000)
--before <date> Pilot history upper date bound
--labels <file> Independent adjudicated label file
--max-files <n> Snapshot file budget (1..10000)
--unit <kind> Prepared evidence: function, hunk, or all
--bca-command <path> Optional pinned bca executable for preparation
--bca-version <text> Exact required bca --version stdout
--comparison <file> Matched retrieval-arm labeled cohort for five-arm audit
--json Machine-readable output (always enabled)
-h, --help display help for command
metrics decisions pilot
Usage: interlinked metrics decisions pilot [options]
Sample real changes and retain unavailable revision counts
Options:
--cwd <path> Project root
--config <file> Explicit decision configuration
--output <file> Write an immutable JSON artifact
--metrics <ids> Comma-separated metric contract IDs
--band <low:high> Uncertainty band; probabilities are reusable
--command <path> ThinkThen executable
--max-records <n> Record budget; excluded records remain unassessed
--plan Print native request plans without inference
--fresh Fresh inference; preserve the previous cached
observation
--base <ref> Git base revision
--revision <ref> Committed revision; default is working tree
--file <paths...> Restrict changed source files
--message <text> Description of a working-tree change
--analysis <kind> atomicity, faithfulness, obligations, compare,
coupling, rework
--claims <file> Literal description claims as a JSON array
--cutoff <date> Train before this date; evaluate at/after it
--after <file> Second retained run for diff
--corpus <file> Snapshot or historical function corpus
--retrieval <method> lexical, embedding, or hybrid
--count <n> Pilot commit budget (1..1000)
--before <date> Pilot history upper date bound
--labels <file> Independent adjudicated label file
--max-files <n> Snapshot file budget (1..10000)
--unit <kind> Prepared evidence: function, hunk, or all
--bca-command <path> Optional pinned bca executable for preparation
--bca-version <text> Exact required bca --version stdout
--comparison <file> Matched retrieval-arm labeled cohort for five-arm audit
--json Machine-readable output (always enabled)
-h, --help display help for command
metrics decisions labels
Usage: interlinked metrics decisions labels [options] <input>
Create a blind independent-review label template from a run
Options:
--cwd <path> Project root
--config <file> Explicit decision configuration
--output <file> Write an immutable JSON artifact
--metrics <ids> Comma-separated metric contract IDs
--band <low:high> Uncertainty band; probabilities are reusable
--command <path> ThinkThen executable
--max-records <n> Record budget; excluded records remain unassessed
--plan Print native request plans without inference
--fresh Fresh inference; preserve the previous cached
observation
--base <ref> Git base revision
--revision <ref> Committed revision; default is working tree
--file <paths...> Restrict changed source files
--message <text> Description of a working-tree change
--analysis <kind> atomicity, faithfulness, obligations, compare,
coupling, rework
--claims <file> Literal description claims as a JSON array
--cutoff <date> Train before this date; evaluate at/after it
--after <file> Second retained run for diff
--corpus <file> Snapshot or historical function corpus
--retrieval <method> lexical, embedding, or hybrid
--count <n> Pilot commit budget (1..1000)
--before <date> Pilot history upper date bound
--labels <file> Independent adjudicated label file
--max-files <n> Snapshot file budget (1..10000)
--unit <kind> Prepared evidence: function, hunk, or all
--bca-command <path> Optional pinned bca executable for preparation
--bca-version <text> Exact required bca --version stdout
--comparison <file> Matched retrieval-arm labeled cohort for five-arm audit
--json Machine-readable output (always enabled)
-h, --help display help for command
metrics decisions join-labels
Usage: interlinked metrics decisions join-labels [options] <input>
Join complete adjudicated labels with saved probabilities
Options:
--cwd <path> Project root
--config <file> Explicit decision configuration
--output <file> Write an immutable JSON artifact
--metrics <ids> Comma-separated metric contract IDs
--band <low:high> Uncertainty band; probabilities are reusable
--command <path> ThinkThen executable
--max-records <n> Record budget; excluded records remain unassessed
--plan Print native request plans without inference
--fresh Fresh inference; preserve the previous cached
observation
--base <ref> Git base revision
--revision <ref> Committed revision; default is working tree
--file <paths...> Restrict changed source files
--message <text> Description of a working-tree change
--analysis <kind> atomicity, faithfulness, obligations, compare,
coupling, rework
--claims <file> Literal description claims as a JSON array
--cutoff <date> Train before this date; evaluate at/after it
--after <file> Second retained run for diff
--corpus <file> Snapshot or historical function corpus
--retrieval <method> lexical, embedding, or hybrid
--count <n> Pilot commit budget (1..1000)
--before <date> Pilot history upper date bound
--labels <file> Independent adjudicated label file
--max-files <n> Snapshot file budget (1..10000)
--unit <kind> Prepared evidence: function, hunk, or all
--bca-command <path> Optional pinned bca executable for preparation
--bca-version <text> Exact required bca --version stdout
--comparison <file> Matched retrieval-arm labeled cohort for five-arm audit
--json Machine-readable output (always enabled)
-h, --help display help for command
metrics decisions probe
Usage: interlinked metrics decisions probe [options] <input>
Run fresh perturbation, repetition, and batch-composition probes
Options:
--cwd <path> Project root
--config <file> Explicit decision configuration
--output <file> Write an immutable JSON artifact
--metrics <ids> Comma-separated metric contract IDs
--band <low:high> Uncertainty band; probabilities are reusable
--command <path> ThinkThen executable
--max-records <n> Record budget; excluded records remain unassessed
--plan Print native request plans without inference
--fresh Fresh inference; preserve the previous cached
observation
--base <ref> Git base revision
--revision <ref> Committed revision; default is working tree
--file <paths...> Restrict changed source files
--message <text> Description of a working-tree change
--analysis <kind> atomicity, faithfulness, obligations, compare,
coupling, rework
--claims <file> Literal description claims as a JSON array
--cutoff <date> Train before this date; evaluate at/after it
--after <file> Second retained run for diff
--corpus <file> Snapshot or historical function corpus
--retrieval <method> lexical, embedding, or hybrid
--count <n> Pilot commit budget (1..1000)
--before <date> Pilot history upper date bound
--labels <file> Independent adjudicated label file
--max-files <n> Snapshot file budget (1..10000)
--unit <kind> Prepared evidence: function, hunk, or all
--bca-command <path> Optional pinned bca executable for preparation
--bca-version <text> Exact required bca --version stdout
--comparison <file> Matched retrieval-arm labeled cohort for five-arm audit
--json Machine-readable output (always enabled)
-h, --help display help for command
metrics decisions measure
Usage: interlinked metrics decisions measure [options] <input>
Measure an evidence population and retain raw probabilities
Options:
--cwd <path> Project root
--config <file> Explicit decision configuration
--output <file> Write an immutable JSON artifact
--metrics <ids> Comma-separated metric contract IDs
--band <low:high> Uncertainty band; probabilities are reusable
--command <path> ThinkThen executable
--max-records <n> Record budget; excluded records remain unassessed
--plan Print native request plans without inference
--fresh Fresh inference; preserve the previous cached
observation
--base <ref> Git base revision
--revision <ref> Committed revision; default is working tree
--file <paths...> Restrict changed source files
--message <text> Description of a working-tree change
--analysis <kind> atomicity, faithfulness, obligations, compare,
coupling, rework
--claims <file> Literal description claims as a JSON array
--cutoff <date> Train before this date; evaluate at/after it
--after <file> Second retained run for diff
--corpus <file> Snapshot or historical function corpus
--retrieval <method> lexical, embedding, or hybrid
--count <n> Pilot commit budget (1..1000)
--before <date> Pilot history upper date bound
--labels <file> Independent adjudicated label file
--max-files <n> Snapshot file budget (1..10000)
--unit <kind> Prepared evidence: function, hunk, or all
--bca-command <path> Optional pinned bca executable for preparation
--bca-version <text> Exact required bca --version stdout
--comparison <file> Matched retrieval-arm labeled cohort for five-arm audit
--json Machine-readable output (always enabled)
-h, --help display help for command
metrics decisions review
Usage: interlinked metrics decisions review [options] <input>
Run relational obligations, atomicity, faithfulness, or comparison
Options:
--cwd <path> Project root
--config <file> Explicit decision configuration
--output <file> Write an immutable JSON artifact
--metrics <ids> Comma-separated metric contract IDs
--band <low:high> Uncertainty band; probabilities are reusable
--command <path> ThinkThen executable
--max-records <n> Record budget; excluded records remain unassessed
--plan Print native request plans without inference
--fresh Fresh inference; preserve the previous cached
observation
--base <ref> Git base revision
--revision <ref> Committed revision; default is working tree
--file <paths...> Restrict changed source files
--message <text> Description of a working-tree change
--analysis <kind> atomicity, faithfulness, obligations, compare,
coupling, rework
--claims <file> Literal description claims as a JSON array
--cutoff <date> Train before this date; evaluate at/after it
--after <file> Second retained run for diff
--corpus <file> Snapshot or historical function corpus
--retrieval <method> lexical, embedding, or hybrid
--count <n> Pilot commit budget (1..1000)
--before <date> Pilot history upper date bound
--labels <file> Independent adjudicated label file
--max-files <n> Snapshot file budget (1..10000)
--unit <kind> Prepared evidence: function, hunk, or all
--bca-command <path> Optional pinned bca executable for preparation
--bca-version <text> Exact required bca --version stdout
--comparison <file> Matched retrieval-arm labeled cohort for five-arm audit
--json Machine-readable output (always enabled)
-h, --help display help for command
metrics decisions embed
Usage: interlinked metrics decisions embed [options] <input>
Embed evidence locally, then unload the model
Options:
--cwd <path> Project root
--config <file> Explicit decision configuration
--output <file> Write an immutable JSON artifact
--metrics <ids> Comma-separated metric contract IDs
--band <low:high> Uncertainty band; probabilities are reusable
--command <path> ThinkThen executable
--max-records <n> Record budget; excluded records remain unassessed
--plan Print native request plans without inference
--fresh Fresh inference; preserve the previous cached
observation
--base <ref> Git base revision
--revision <ref> Committed revision; default is working tree
--file <paths...> Restrict changed source files
--message <text> Description of a working-tree change
--analysis <kind> atomicity, faithfulness, obligations, compare,
coupling, rework
--claims <file> Literal description claims as a JSON array
--cutoff <date> Train before this date; evaluate at/after it
--after <file> Second retained run for diff
--corpus <file> Snapshot or historical function corpus
--retrieval <method> lexical, embedding, or hybrid
--count <n> Pilot commit budget (1..1000)
--before <date> Pilot history upper date bound
--labels <file> Independent adjudicated label file
--max-files <n> Snapshot file budget (1..10000)
--unit <kind> Prepared evidence: function, hunk, or all
--bca-command <path> Optional pinned bca executable for preparation
--bca-version <text> Exact required bca --version stdout
--comparison <file> Matched retrieval-arm labeled cohort for five-arm audit
--json Machine-readable output (always enabled)
-h, --help display help for command
metrics decisions geometry
Usage: interlinked metrics decisions geometry [options] <input>
Measure provisional embedding churn, dispersion, and nearest-neighbor outliers
Options:
--cwd <path> Project root
--config <file> Explicit decision configuration
--output <file> Write an immutable JSON artifact
--metrics <ids> Comma-separated metric contract IDs
--band <low:high> Uncertainty band; probabilities are reusable
--command <path> ThinkThen executable
--max-records <n> Record budget; excluded records remain unassessed
--plan Print native request plans without inference
--fresh Fresh inference; preserve the previous cached
observation
--base <ref> Git base revision
--revision <ref> Committed revision; default is working tree
--file <paths...> Restrict changed source files
--message <text> Description of a working-tree change
--analysis <kind> atomicity, faithfulness, obligations, compare,
coupling, rework
--claims <file> Literal description claims as a JSON array
--cutoff <date> Train before this date; evaluate at/after it
--after <file> Second retained run for diff
--corpus <file> Snapshot or historical function corpus
--retrieval <method> lexical, embedding, or hybrid
--count <n> Pilot commit budget (1..1000)
--before <date> Pilot history upper date bound
--labels <file> Independent adjudicated label file
--max-files <n> Snapshot file budget (1..10000)
--unit <kind> Prepared evidence: function, hunk, or all
--bca-command <path> Optional pinned bca executable for preparation
--bca-version <text> Exact required bca --version stdout
--comparison <file> Matched retrieval-arm labeled cohort for five-arm audit
--json Machine-readable output (always enabled)
-h, --help display help for command
metrics decisions audit
Usage: interlinked metrics decisions audit [options] <input>
Evaluate a full labeled cohort with a temporal split and baselines
Options:
--cwd <path> Project root
--config <file> Explicit decision configuration
--output <file> Write an immutable JSON artifact
--metrics <ids> Comma-separated metric contract IDs
--band <low:high> Uncertainty band; probabilities are reusable
--command <path> ThinkThen executable
--max-records <n> Record budget; excluded records remain unassessed
--plan Print native request plans without inference
--fresh Fresh inference; preserve the previous cached
observation
--base <ref> Git base revision
--revision <ref> Committed revision; default is working tree
--file <paths...> Restrict changed source files
--message <text> Description of a working-tree change
--analysis <kind> atomicity, faithfulness, obligations, compare,
coupling, rework
--claims <file> Literal description claims as a JSON array
--cutoff <date> Train before this date; evaluate at/after it
--after <file> Second retained run for diff
--corpus <file> Snapshot or historical function corpus
--retrieval <method> lexical, embedding, or hybrid
--count <n> Pilot commit budget (1..1000)
--before <date> Pilot history upper date bound
--labels <file> Independent adjudicated label file
--max-files <n> Snapshot file budget (1..10000)
--unit <kind> Prepared evidence: function, hunk, or all
--bca-command <path> Optional pinned bca executable for preparation
--bca-version <text> Exact required bca --version stdout
--comparison <file> Matched retrieval-arm labeled cohort for five-arm audit
--json Machine-readable output (always enabled)
-h, --help display help for command
metrics decisions replay
Usage: interlinked metrics decisions replay [options] <input>
Reinterpret a retained run at a new uncertainty band, offline
Options:
--cwd <path> Project root
--config <file> Explicit decision configuration
--output <file> Write an immutable JSON artifact
--metrics <ids> Comma-separated metric contract IDs
--band <low:high> Uncertainty band; probabilities are reusable
--command <path> ThinkThen executable
--max-records <n> Record budget; excluded records remain unassessed
--plan Print native request plans without inference
--fresh Fresh inference; preserve the previous cached
observation
--base <ref> Git base revision
--revision <ref> Committed revision; default is working tree
--file <paths...> Restrict changed source files
--message <text> Description of a working-tree change
--analysis <kind> atomicity, faithfulness, obligations, compare,
coupling, rework
--claims <file> Literal description claims as a JSON array
--cutoff <date> Train before this date; evaluate at/after it
--after <file> Second retained run for diff
--corpus <file> Snapshot or historical function corpus
--retrieval <method> lexical, embedding, or hybrid
--count <n> Pilot commit budget (1..1000)
--before <date> Pilot history upper date bound
--labels <file> Independent adjudicated label file
--max-files <n> Snapshot file budget (1..10000)
--unit <kind> Prepared evidence: function, hunk, or all
--bca-command <path> Optional pinned bca executable for preparation
--bca-version <text> Exact required bca --version stdout
--comparison <file> Matched retrieval-arm labeled cohort for five-arm audit
--json Machine-readable output (always enabled)
-h, --help display help for command
metrics decisions diff
Usage: interlinked metrics decisions diff [options] <input>
Compare retained answers and report model/question drift
Options:
--cwd <path> Project root
--config <file> Explicit decision configuration
--output <file> Write an immutable JSON artifact
--metrics <ids> Comma-separated metric contract IDs
--band <low:high> Uncertainty band; probabilities are reusable
--command <path> ThinkThen executable
--max-records <n> Record budget; excluded records remain unassessed
--plan Print native request plans without inference
--fresh Fresh inference; preserve the previous cached
observation
--base <ref> Git base revision
--revision <ref> Committed revision; default is working tree
--file <paths...> Restrict changed source files
--message <text> Description of a working-tree change
--analysis <kind> atomicity, faithfulness, obligations, compare,
coupling, rework
--claims <file> Literal description claims as a JSON array
--cutoff <date> Train before this date; evaluate at/after it
--after <file> Second retained run for diff
--corpus <file> Snapshot or historical function corpus
--retrieval <method> lexical, embedding, or hybrid
--count <n> Pilot commit budget (1..1000)
--before <date> Pilot history upper date bound
--labels <file> Independent adjudicated label file
--max-files <n> Snapshot file budget (1..10000)
--unit <kind> Prepared evidence: function, hunk, or all
--bca-command <path> Optional pinned bca executable for preparation
--bca-version <text> Exact required bca --version stdout
--comparison <file> Matched retrieval-arm labeled cohort for five-arm audit
--json Machine-readable output (always enabled)
-h, --help display help for command
Coverage
Usage: interlinked coverage [options] [command]
Per-file coverage ratchet — fails on any file whose coverage drops
Options:
-h, --help display help for command
Commands:
move [options] [old] [new] Record an e2e boundary move, preserving every
floor
retire [options] <path> Retire an e2e floor whose boundary file is gone
or erased
status [options] Show quality coverage baselines by lane (distinct
from harness filesystem coverage)
check [options] Compare current coverage against the baseline.
Per-file drops are ADVISORY (exit 0) unless
--strict is passed
metrics [options] Per-metric distribution (lines / statements /
functions / branches) over the merged coverage
report: measured, at 100%, under a threshold,
p50/p90, lowest files
baseline [options] Show the current coverage baseline
help [command] display help for command
coverage check
Usage: interlinked coverage check [options]
Compare current coverage against the baseline. Per-file drops are ADVISORY
(exit 0) unless --strict is passed
Options:
--report <path> Path to one coverage report (LCOV .info or istanbul
JSON). Default: merge every discovered coverage
report
--changed-files <list> Comma-separated repo-relative paths; only report
drops for these files
--update-baseline Persist the current coverage as the new baseline
--from-index Measure through the coverage index: re-run only the
test shards whose inputs changed, keep every other
shard and the full denominator (no verdict ⇒ exit 75)
--index-store <root> Checkout whose .interlinked/coverage-index/ holds the
index (default: the project root); the pre-push
export names its source checkout
--timeout <ms> Index measurement budget in milliseconds (default
3600000)
--workers <n> Requested worker cap for the index's instrumented
capture, clamped by the host governor (default 1)
--lane <name> Named coverage lane: e2e (isolated strict policy)
--init-baseline Initialize an absent e2e baseline from a passing
measured run
--require-measured Fail on a partial or unmeasured report
--base <ref> E2e comparison base (default: HEAD; CI supplies its
event base)
--map <old=new> Apply a mapping in the measured e2e transaction;
repeatable
--strict exit non-zero on any per-file drop (default:
advisory)
--cwd <path> Project root (default: current directory)
--json Machine-readable output
-h, --help display help for command
coverage baseline
Usage: interlinked coverage baseline [options]
Show the current coverage baseline
Options:
--lane <name> Named coverage lane: e2e
--cwd <path> Project root
--json Machine-readable output
-h, --help display help for command
coverage metrics
Usage: interlinked coverage metrics [options]
Per-metric distribution (lines / statements / functions / branches) over the
merged coverage report: measured, at 100%, under a threshold, p50/p90, lowest
files
Options:
--report <path> Path to one coverage report (LCOV .info or istanbul JSON).
Default: merge every discovered report
--metric <name> Only this metric: lines | statements | functions | branches
--under <pct> Threshold for the "under N%" count (default 90)
--top <n> How many lowest files to list per metric (default 10)
--cwd <path> Project root (default: current directory)
--json Machine-readable output
-h, --help display help for command
coverage move
Usage: interlinked coverage move [options] [old] [new]
Record an e2e boundary move, preserving every floor
Options:
--lane <name> Required: e2e
--map <old=new> Atomic mapping; repeat for a chain or swap
--base <ref> Comparison base (default: HEAD)
--cwd <path> Project root
--json Machine-readable output
-h, --help display help for command
coverage retire
Usage: interlinked coverage retire [options] <path>
Retire an e2e floor whose boundary file is gone or erased
Options:
--lane <name> Required: e2e
--base <ref> Comparison base (default: HEAD)
--cwd <path> Project root
--json Machine-readable output
-h, --help display help for command
coverage status
Usage: interlinked coverage status [options]
Show quality coverage baselines by lane (distinct from harness filesystem
coverage)
Options:
--cwd <path> Project root
--json Machine-readable output
-h, --help display help for command
E2e
Usage: interlinked e2e [options] [command]
Interlinked's OWN hook/daemon boundary tests (self-test lane); host projects
use `interlinked tests e2e`
Options:
-h, --help display help for command
Commands:
scaffold [options] <name> Create a fixture-backed e2e test with a failing
contract assertion (Interlinked checkout only;
developer preset elsewhere)
help [command] display help for command
e2e scaffold
Usage: interlinked e2e scaffold [options] <name>
Create a fixture-backed e2e test with a failing contract assertion (Interlinked
checkout only; developer preset elsewhere)
Options:
--event <event> PreToolUse, PostToolUse or Stop (default: "PreToolUse")
--tool <tool> Native tool name (default: "Edit")
--cwd <path> Project root
--dry-run Print the test without writing it
--developer-preset Write the Interlinked-private preset outside the
Interlinked checkout (never emitted into a host project
by default)
-h, --help display help for command
Simplify
Usage: interlinked simplify [options] [command]
Simplification evidence, explicit recording, and Agent CI handoff preparation
Options:
-h, --help display help for command
Commands:
scan [options] Compose deterministic local simplification evidence for the
repository
review [options] Review changed, staged, or explicit git-range files using
full repository context
audit [options] Run the local deterministic repository audit; never invokes
an LLM
status [options] Show locally recorded simplification runs and common-corpus
materialization
help [command] display help for command
simplify scan
Usage: interlinked simplify scan [options]
Compose deterministic local simplification evidence for the repository
Options:
--cwd <path> Project root (default: current directory)
--record Persist a local run receipt and upsert findings into the common
corpus
--json Canonical machine-readable report with evidence and coverage
receipts
-h, --help display help for command
simplify review
Usage: interlinked simplify review [options]
Review changed, staged, or explicit git-range files using full repository
context
Options:
--diagnostics <profile> Include iterative diagnostic candidates: js-ts or
python (explicit full census)
--changed Review tracked and untracked worktree changes
(default)
--staged Review only paths staged in the git index
--range <base..head> Review paths changed by an explicit validated git
range
--deep-handoff Include a portable, not-submitted Agent CI
deep-review request
--cwd <path> Project root (default: current directory)
--record Persist a local run receipt and upsert findings into
the common corpus
--json Canonical machine-readable report with evidence and
coverage receipts
-h, --help display help for command
simplify audit
Usage: interlinked simplify audit [options]
Run the local deterministic repository audit; never invokes an LLM
Options:
--deep-handoff Include a portable, not-submitted Agent CI deep-review
request
--cwd <path> Project root (default: current directory)
--record Persist a local run receipt and upsert findings into the
common corpus
--json Canonical machine-readable report with evidence and coverage
receipts
-h, --help display help for command
simplify status
Usage: interlinked simplify status [options]
Show locally recorded simplification runs and common-corpus materialization
Options:
--cwd <path> Project root (default: current directory)
--json Machine-readable local recorded-run view
-h, --help display help for command
Tests
Usage: interlinked tests [options] [command]
Explain and run affected tests with bounded scheduling and snapshot validation
Options:
-h, --help display help for command
Commands:
contracts Inspect provenance or explicitly execute
portable behavioral contracts
e2e Project-aware end-to-end scenarios: declared
policy, supervised runs, receipts and one
qualification verdict
review [options] [paths...] Review changed source, behavioral obligations
and bounded simplification candidates without
running tests
readiness [options] <language> Inspect test prerequisites without
installing; Python also reports approved
provisioning argv
suite [options] <language> Run a bounded project suite: typescript,
javascript, python, rust or go
plan [options] [paths...] Explain the union of affected tests without
running assertions
run [options] [paths...] Run pending and affected tests; retain newer
edits for another run
status [options] [paths...] Show pending requests and the last observed
test job
help [command] display help for command
tests plan
Usage: interlinked tests plan [options] [paths...]
Explain the union of affected tests without running assertions
Options:
--cwd <path> Project root
--base <ref> Compare tracked changes against this revision
(default: "HEAD")
--all Run the full suite for reconciliation
--timeout <ms> Total planning, admission and execution budget
(default: "60000")
--workers <count> Maximum workers, also bounded by available memory
(default: "2")
--coverage Collect coverage (json-summary) into the run
directory; part of the check identity
--coverage-reporter <path> Extra vitest reporter module for coverage runs
(repeatable)
--receipt-store <dir> Read and write passing-run receipts and run
artifacts here instead of
<cwd>/.interlinked/test-runs
--artifacts-out <dir> Copy the certified run's artifacts (fresh or
reused) into this directory after a passing run
--json Machine-readable test plan or execution
--details Show complete test-plan and pending-input
listings
--static-selection Let the effect classifier omit unaffected tests
and reuse passing results (off by default: it
cannot prove either safe)
-h, --help display help for command
tests run
Usage: interlinked tests run [options] [paths...]
Run pending and affected tests; retain newer edits for another run
Options:
--cwd <path> Project root
--base <ref> Compare tracked changes against this revision
(default: "HEAD")
--all Run the full suite for reconciliation
--timeout <ms> Total planning, admission and execution budget
(default: "60000")
--workers <count> Maximum workers, also bounded by available memory
(default: "2")
--coverage Collect coverage (json-summary) into the run
directory; part of the check identity
--coverage-reporter <path> Extra vitest reporter module for coverage runs
(repeatable)
--receipt-store <dir> Read and write passing-run receipts and run
artifacts here instead of
<cwd>/.interlinked/test-runs
--artifacts-out <dir> Copy the certified run's artifacts (fresh or
reused) into this directory after a passing run
--json Machine-readable test plan or execution
--details Show complete test-plan and pending-input
listings
--static-selection Let the effect classifier omit unaffected tests
and reuse passing results (off by default: it
cannot prove either safe)
-h, --help display help for command
tests status
Usage: interlinked tests status [options] [paths...]
Show pending requests and the last observed test job
Options:
--cwd <path> Project root
--base <ref> Compare tracked changes against this revision
(default: "HEAD")
--all Run the full suite for reconciliation
--timeout <ms> Total planning, admission and execution budget
(default: "60000")
--workers <count> Maximum workers, also bounded by available memory
(default: "2")
--coverage Collect coverage (json-summary) into the run
directory; part of the check identity
--coverage-reporter <path> Extra vitest reporter module for coverage runs
(repeatable)
--receipt-store <dir> Read and write passing-run receipts and run
artifacts here instead of
<cwd>/.interlinked/test-runs
--artifacts-out <dir> Copy the certified run's artifacts (fresh or
reused) into this directory after a passing run
--json Machine-readable test plan or execution
--details Show complete test-plan and pending-input
listings
--static-selection Let the effect classifier omit unaffected tests
and reuse passing results (off by default: it
cannot prove either safe)
-h, --help display help for command
tests suite
Usage: interlinked tests suite [options] <language>
Run a bounded project suite: typescript, javascript, python, rust or go
Options:
--cwd <path> Project root
--timeout <ms> Total admission and execution budget (default: "60000")
--json Machine-readable execution evidence
-h, --help display help for command
tests readiness
Usage: interlinked tests readiness [options] <language>
Inspect test prerequisites without installing; Python also reports approved
provisioning argv
Options:
--cwd <path> Project root
--json Machine-readable readiness and missing prerequisites
-h, --help display help for command
tests review
Usage: interlinked tests review [options] [paths...]
Review changed source, behavioral obligations and bounded simplification
candidates without running tests
Options:
--cwd <path> Project root
--base <ref> Revision for automatic changed-file discovery (default: "HEAD")
--json Machine-readable review scope and missing evidence
-h, --help display help for command
tests contracts
Usage: interlinked tests contracts [options] [command]
Inspect provenance or explicitly execute portable behavioral contracts
Options:
-h, --help display help for command
Commands:
import [options] <source> Print proposed cases from explicit json
interlinked-contract fences; never execute
inspect [options] Inspect provenance and configured acceptance
without execution
run [options] Execute declared runners and retain per-case
receipts
help [command] display help for command
tests contracts import
Usage: interlinked tests contracts import [options] <source>
Print proposed cases from explicit json interlinked-contract fences; never
execute
Options:
--cwd <path> Project root
--json Machine-readable proposed manifest
-h, --help display help for command
tests contracts inspect
Usage: interlinked tests contracts inspect [options]
Inspect provenance and configured acceptance without execution
Options:
--cwd <path> Project root
--file <path> Project-relative contract manifest (default:
".interlinked/behavioral-contracts.json")
--previous <path> Prior manifest whose cases also run against current
implementation
--timeout <ms> Total admission and execution budget (default: "60000")
--json Machine-readable contract evidence
-h, --help display help for command
tests contracts run
Usage: interlinked tests contracts run [options]
Execute declared runners and retain per-case receipts
Options:
--cwd <path> Project root
--file <path> Project-relative contract manifest (default:
".interlinked/behavioral-contracts.json")
--previous <path> Prior manifest whose cases also run against current
implementation
--timeout <ms> Total admission and execution budget (default: "60000")
--json Machine-readable contract evidence
-h, --help display help for command
Allowlist
Usage: interlinked allowlist [options] [command]
Manage the supply-chain package allowlist (.interlinked/package-allowlist.json)
Options:
-h, --help display help for command
Commands:
add [options] <ecosystem> <package> Approve a registry package (ecosystem: npm | pypi | cargo | rubygems | go)
remove [options] <ecosystem> <package> Un-approve a previously-approved package
propose [options] <ecosystem> <package> Record a dependency request without changing approval or installing it
list [options] Show approved packages and snapshots
snapshot [options] Hash current manifest+lockfile state, store as an approved snapshot
verify [options] Show manifest deps not on the allowlist
help [command] display help for command
allowlist add
Usage: interlinked allowlist add [options] <ecosystem> <package>
Approve a registry package (ecosystem: npm | pypi | cargo | rubygems | go)
Options:
--by <name> Approver name (required)
--reason <text> Why this package is approved
--version-range <range> Optional semver/PEP-440 range constraint
--force Override the admission screens (typosquat refusal,
non-allowlisted license, open OSV advisories)
--cwd <path> Project root (default: current directory)
-h, --help display help for command
allowlist propose
Usage: interlinked allowlist propose [options] <ecosystem> <package>
Record a dependency request without changing approval or installing it
Options:
--package-version <version> Requested package version
--reason <text> Why the dependency is needed
--cwd <path> Project root
--json Machine-readable proposal
-h, --help display help for command
allowlist remove
Usage: interlinked allowlist remove [options] <ecosystem> <package>
Un-approve a previously-approved package
Options:
--cwd <path> Project root
-h, --help display help for command
allowlist list
Usage: interlinked allowlist list [options]
Show approved packages and snapshots
Options:
--ecosystem <name> Filter by ecosystem
--json Machine-readable output
--cwd <path> Project root
-h, --help display help for command
allowlist snapshot
Usage: interlinked allowlist snapshot [options]
Hash current manifest+lockfile state, store as an approved snapshot
Options:
--by <name> Approver name (required)
--reason <text> Why this state is approved
--lockfile <name> Snapshot a specific file only (e.g. package-lock.json)
--cwd <path> Project root
-h, --help display help for command
allowlist verify
Usage: interlinked allowlist verify [options]
Show manifest deps not on the allowlist
Options:
--cwd <path> Project root
-h, --help display help for command
Harness
Usage: interlinked harness [options] [command]
Local harness server: guard evaluation, auto-reservations, agent lifecycle
Options:
-h, --help display help for command
Commands:
start [options] Start the harness server (background daemon by
default)
stop [options] Stop the harness server
restart [options] Stop and restart the harness server (picks up
config changes)
status [options] Show harness status, loaded rules, and active
agents
health [options] Check health: recurrence, finding outcomes,
compatible rechecks and missing evidence
corpus Inspect and refresh versioned detector corpus
evidence
explain [options] <topic> Explain implementation, local policy, client
support and observed evidence separately
checks [options] Show the authoritative check inventory —
per-family counts + total (static; no daemon
needed)
capabilities [options] Inspect runtime hook profiles, selected events,
installation and observation evidence
coverage Inspect and review daemon-owned filesystem
observations
taint Session sensitivity label: inspect which source
set it, or clear it (operator)
test [options] [command] Fire a synthetic PreToolUse event at the running
harness and show its decision. Default: a Bash
command. Use --write/--edit to test a file
operation.
reap [options] List (default) or kill orphan harness daemons.
--force to SIGTERM. --all also targets the active
daemon.
clean [options] Remove stale harness.sock + harness.pid (refuses
if a daemon is running)
mode [options] [name] Show or switch the operational tier
(budget|quality|ci) — drives
HARNESS_POST_TIMEOUT_MS
latency [options] Show per-event latency report from
.interlinked/logs/latency.jsonl
help [command] display help for command
harness start
Usage: interlinked harness start [options]
Start the harness server (background daemon by default)
Options:
--no-daemon Run in foreground instead of background
--protocol <mode> Socket protocol: raw, framed, or dual (default: "dual")
--session-id <id> Framed socket session id (default: "default")
--verbose Verbose logging
--json Machine-readable output
-h, --help display help for command
harness stop
Usage: interlinked harness stop [options]
Stop the harness server
Options:
--json Machine-readable output
-h, --help display help for command
harness restart
Usage: interlinked harness restart [options]
Stop and restart the harness server (picks up config changes)
Options:
--no-daemon Run in foreground instead of daemon
--protocol <mode> Socket protocol: raw, framed, or dual (default: "dual")
--session-id <id> Framed socket session id (default: "default")
--verbose Verbose output
--json Machine-readable output
-h, --help display help for command
harness status
Usage: interlinked harness status [options]
Show harness status, loaded rules, and active agents
Options:
--json Machine-readable output
-h, --help display help for command
harness checks
Usage: interlinked harness checks [options]
Show the authoritative check inventory — per-family counts + total (static; no
daemon needed)
Options:
--json Machine-readable output
--short One-line summary
--full Include each count's authoritative source
-h, --help display help for command
harness capabilities
Usage: interlinked harness capabilities [options]
Inspect runtime hook profiles, selected events, installation and observation
evidence
Options:
--json Full capability catalog and coverage identities
-h, --help display help for command
harness coverage
Usage: interlinked harness coverage [options] [command]
Inspect and review daemon-owned filesystem observations
Options:
-h, --help display help for command
Commands:
status [options]
verify [options] Check pending file versions and retain exact-version evidence and findings
repair-history [options] Set aside a history log shorter than its committed pointer and restart the history; keeps state and pending obligations (offline; restart the daemon after)
accept-policy [options] <digest> Explicitly accept the exact current protected-policy digest
acknowledge [options] <id> <generation> <identity> <evidence> Record manual review evidence for an exact pending file version; does not claim automated checks
help [command] display help for command
harness coverage status
Usage: interlinked harness coverage status [options]
Options:
--json Machine-readable output
--progress Read cached recovery progress without rescanning files or
returning receipt history
-h, --help display help for command
harness coverage verify
Usage: interlinked harness coverage verify [options]
Check pending file versions and retain exact-version evidence and findings
Options:
--json Machine-readable output
--no-wait Start the daemon verification job and return immediately
-h, --help display help for command
harness coverage acknowledge
Usage: interlinked harness coverage acknowledge [options] <id> <generation> <identity> <evidence>
Record manual review evidence for an exact pending file version; does not claim
automated checks
Options:
--json Machine-readable output
-h, --help display help for command
harness coverage accept-policy
Usage: interlinked harness coverage accept-policy [options] <digest>
Explicitly accept the exact current protected-policy digest
Options:
--json Machine-readable output
-h, --help display help for command
harness test
Usage: interlinked harness test [options] [command]
Fire a synthetic PreToolUse event at the running harness and show its decision.
Default: a Bash command. Use --write/--edit to test a file operation.
Options:
--tool <name> Tool name to simulate for the positional command
(default: "Bash")
--write <file> Simulate a Write to <file> (pair with --from-file or
--stdin)
--from-file <path> Read the proposed Write content from <path>
--stdin Read the proposed Write content from stdin
--edit <file> Simulate an Edit to <file> (pair with --old and --new)
--old <str> old_string for --edit
--new <str> new_string for --edit
--json Machine-readable output
-h, --help display help for command
Checkpoint
Usage: interlinked checkpoint [options] [command] [message]
Git checkpoint management
Arguments:
message Checkpoint message
Options:
--agent <name> Agent name
--json Machine-readable output
-h, --help display help for command
Commands:
list [options] List checkpoints
show [options] <id> Show checkpoint details
compare [options] <id1> <id2> Diff two checkpoints
prune [options] Remove old checkpoints
archive [options] Archive old stash checkpoints to metadata-only
checkpoint list
Usage: interlinked checkpoint list [options]
List checkpoints
Options:
--agent <name> Filter by agent
--since <duration> e.g. 1h, 1d
--limit <n> Max entries
--json Machine-readable output
-h, --help display help for command
checkpoint show
Usage: interlinked checkpoint show [options] <id>
Show checkpoint details
Options:
--json Machine-readable output
-h, --help display help for command
checkpoint compare
Usage: interlinked checkpoint compare [options] <id1> <id2>
Diff two checkpoints
Options:
--json Machine-readable output
-h, --help display help for command
checkpoint prune
Usage: interlinked checkpoint prune [options]
Remove old checkpoints
Options:
--older-than <days> Remove checkpoints older than N days
--keep-latest <n> Keep the N most recent checkpoints
--json Machine-readable output
-h, --help display help for command
checkpoint archive
Usage: interlinked checkpoint archive [options]
Archive old stash checkpoints to metadata-only
Options:
--json Machine-readable output
-h, --help display help for command
Tasks
Usage: interlinked tasks [options] [command]
Task management via the server
Options:
-h, --help display help for command
Commands:
list [options] List tasks
create [options] <title> Create a new task
show [options] <id> Show task detail
claim [options] <id> Claim a task
complete [options] <id> Mark a task as complete
help [command] display help for command
tasks list
Usage: interlinked tasks list [options]
List tasks
Options:
--status <status> Filter by status
--assignee <name> Filter by assignee
--priority <level> Filter by priority
--limit <n> Max entries
--json Machine-readable output
--short One-line summary
--full Detailed output
-h, --help display help for command
tasks create
Usage: interlinked tasks create [options] <title>
Create a new task
Options:
--description <text> Task description
--assignee <name> Assign to agent
--priority <level> Task priority
--json Machine-readable output
-h, --help display help for command
tasks show
Usage: interlinked tasks show [options] <id>
Show task detail
Options:
--json Machine-readable output
-h, --help display help for command
tasks claim
Usage: interlinked tasks claim [options] <id>
Claim a task
Options:
--json Machine-readable output
-h, --help display help for command
tasks complete
Usage: interlinked tasks complete [options] <id>
Mark a task as complete
Options:
--json Machine-readable output
-h, --help display help for command
Workspace
Usage: interlinked workspace [options] [command]
Registry workspace management (ws_ IDs)
Options:
-h, --help display help for command
Commands:
list [options] Show workspaces
switch <id> Change active workspace
help [command] display help for command
workspace list
Usage: interlinked workspace list [options]
Show workspaces
Options:
--json Machine-readable output
-h, --help display help for command
workspace switch
Usage: interlinked workspace switch [options] <id>
Change active workspace
Options:
-h, --help display help for command
Index
Usage: interlinked index [options] [command]
Manage the trigram search index for grep acceleration
Options:
-h, --help display help for command
Commands:
build [options] Build a full trigram index from the current
codebase
update [options] Incrementally update the index from git changes
status [options] Show index status and statistics
query [options] <pattern> Query the index for candidate files (debug tool)
help [command] display help for command
index build
Usage: interlinked index build [options]
Build a full trigram index from the current codebase
Options:
--cwd <path> Working directory (default:
"<cwd>")
--max-file-size <bytes> Skip files larger than this (default: "1048576")
--stop-threshold <ratio> Stop trigram threshold (0-1) (default: "0.4")
-h, --help display help for command
index update
Usage: interlinked index update [options]
Incrementally update the index from git changes
Options:
--cwd <path> Working directory (default:
"<cwd>")
-h, --help display help for command
index status
Usage: interlinked index status [options]
Show index status and statistics
Options:
--cwd <path> Working directory (default:
"<cwd>")
--json Output as JSON
-h, --help display help for command
index query
Usage: interlinked index query [options] <pattern>
Query the index for candidate files (debug tool)
Options:
--cwd <path> Working directory (default:
"<cwd>")
--regex Treat pattern as regex (default: false)
-h, --help display help for command
Trace
Usage: interlinked trace [options] [command]
Agent trace export/import
Options:
-h, --help display help for command
Commands:
export [options] Export local activity as agent trace
import [options] <file> Import trace file into local activity
help [command] display help for command
trace export
Usage: interlinked trace export [options]
Export local activity as agent trace
Options:
--since <duration> e.g. 1h, 1d
--agent <name> Filter by agent
--output <file> Output file (default: stdout)
--format <fmt> Format: json (default) or jsonl
--json Machine-readable output
-h, --help display help for command
trace import
Usage: interlinked trace import [options] <file>
Import trace file into local activity
Options:
--json Machine-readable output
-h, --help display help for command
Guard
Usage: interlinked guard [options] [command]
File reservation enforcement via git hooks
Options:
-h, --help display help for command
Commands:
install [options] Install pre-commit hook for reservation checks
check [options] Check files against active reservations
status [options] Show guard configuration and hook status
uninstall [options] Remove guard hooks
help [command] display help for command
guard install
Usage: interlinked guard install [options]
Install pre-commit hook for reservation checks
Options:
--mode <mode> warn (default) or block (default: "warn")
--pre-push Also install pre-push hook
--json Machine-readable output
-h, --help display help for command
guard check
Usage: interlinked guard check [options]
Check files against active reservations
Options:
--files <paths...> File paths to check (default: staged files)
--json Machine-readable output
-h, --help display help for command
guard status
Usage: interlinked guard status [options]
Show guard configuration and hook status
Options:
--json Machine-readable output
-h, --help display help for command
guard uninstall
Usage: interlinked guard uninstall [options]
Remove guard hooks
Options:
--json Machine-readable output
-h, --help display help for command
Git
Usage: interlinked git [options] [command]
Git bridge: metadata, trailers, and notes
Options:
-h, --help display help for command
Commands:
context [options] Show Interlinked metadata for HEAD or a commit
link-checkpoint [options] Link a server checkpoint to a git commit
help [command] display help for command
git context
Usage: interlinked git context [options]
Show Interlinked metadata for HEAD or a commit
Options:
--commit <sha> Specific commit (default: HEAD)
--json Machine-readable output
-h, --help display help for command
git link-checkpoint
Usage: interlinked git link-checkpoint [options]
Link a server checkpoint to a git commit
Options:
--checkpoint <id> Checkpoint ID (default: latest)
--commit <sha> Commit SHA (default: HEAD)
--apply Apply trailers and notes to HEAD (amends commit)
--json Machine-readable output
-h, --help display help for command
Other Commands
enable
Usage: interlinked enable [options]
Install hooks + create .interlinked/ config
Options:
--server <url> Server URL
--agent <name> Default agent name
--clients <list> Comma-separated client list
(claude,copilot,gemini,codex,cursor,opencode,pi)
--sync-mode <mode> Sync mode: realtime (default), local, manual
--data-dir <path> Override data directory for activity logs and sessions
--dry-run Show what would change without modifying files
--structure <mode> Scaffold structure manifests: minimal, standard, strict
-h, --help display help for command
disable
Usage: interlinked disable [options]
Stand the harness down for this project (recorded); --uninstall to remove hooks
+ config
Options:
--team Write a committed marker (shared, shows in PR diffs)
instead of a personal one
--reason <text> Why the harness is being stood down (recorded in the
audit log)
--until <duration> Auto-expire the stand-down (e.g. 30m, 2h, 1d)
--by <name> Who is standing it down (defaults to agent name / $USER)
--uninstall Destructive teardown: remove hooks from all clients and
delete .interlinked/
--keep-config With --uninstall, preserve .interlinked/ config files
-h, --help display help for command
status
Usage: interlinked status [options]
Dashboard: local sessions, recent activity, sync status
Options:
--short One-line summary
--full Per-session detail with tools and files
--json Machine-readable output
--watch [seconds] Auto-refresh interval (default 10s)
-h, --help display help for command
activity
Usage: interlinked activity [options]
Recent activity feed
Options:
--agent <name> Filter by agent
--limit <n> Max entries (default: "30")
--since <duration> e.g. 1h, 30m
--json Machine-readable output
-h, --help display help for command
doctor
Usage: interlinked doctor [options]
Diagnose issues (local + server checks)
Options:
--fix Auto-fix what's possible
--json Machine-readable output
-h, --help display help for command
clean
Usage: interlinked clean [options]
Remove stale data
Options:
--dry-run Show what would be removed (default)
--force Actually delete
--json Machine-readable output
-h, --help display help for command
login
Usage: interlinked login [options]
Authenticate with the server (opens browser)
Options:
--server <url> Server URL
--token <token> Manual token for CI/headless use
-h, --help display help for command
sync
Usage: interlinked sync [options]
Push locally-buffered events to the server
Options:
--dry-run Show what would be synced without sending
--limit <n> Max events to sync
--json Machine-readable output
-h, --help display help for command
send
Usage: interlinked send [options] <to> [message]
Send a message to an agent
Options:
--file <path> Send file contents as message body
--importance <level> Message importance: normal, urgent
--json Machine-readable output
-h, --help display help for command
inbox
Usage: interlinked inbox [options]
Show recent messages from the server
Options:
--all Show all messages (default: unread only)
--agent <name> Filter by recipient agent
--limit <n> Max entries
--since <duration> e.g. 1h, 30m
--json Machine-readable output
--short One-line summary
--full Detailed output
-h, --help display help for command
watch
Usage: interlinked watch [options]
Monitor server for pending work (messages, tasks, agents)
Options:
--interval <seconds> Refresh interval (default 10s)
--short One-line summary
--json Machine-readable output
-h, --help display help for command