Repository navigation
docs(local): document Pi front-blocking and retry boundaries - #1169
Open
shengzhaowei wants to merge 1 commit into
Open
shengzhaowei wants to merge 1 commit into
shengzhaowei wants to merge 1 commit into
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What changed
local/skills/devflow/adapters/pi/adapter.mdalready documented that a front-blockingsubagentcall cannot be reused and that retry must re-dispatch a new instance, but it said nothing about what the coordinator can and cannot do during the blocking call, or about the loop that runs inside it. This PR adds a "前台阻塞与重试边界" section: (1) During a front-blocking call the coordinator cannot steer mid-flight: it cannot append instructions, change the prompt, or stop the sub-session while it is running. It must wait for the return, verify the artifacts, and re-dispatch a new instance if more work is needed. (2)runLoopis a two-layer while loop with no hard-coded step cap, so the step cap must not be treated as a safety net. Stage prompts must converge first — goal, artifacts, and acceptance criteria written clearly in one pass — and long context must be summarised before dispatch, not handed to the sub-session unconverged. (3) Retry is cross-referenced to capability-map item 5: re-dispatch a new instance and re-register it; never reuse a finished Mission id. A regression test asserts the key sentences exist in the adapter doc, following the existing CodeBuddy adapter test pattern.Scope
Portable (checked), Documentation only (checked); all other scope boxes unchecked.
Compatibility
No behavior change. No migration, no version bump, no Classic bundle change. Only the Pi adapter text and its regression test change; other adapters are unaffected.
Validation
Run locally on 2026-11-21, base d58a85a: validate_config OK (adapters=7 logical_roles=10 host_agents=6 artifacts=11); unittest 31 tests OK (incl. test_pi_adapter_documents_front_blocking_and_retry_boundaries); validate.sh [1/6]-[6/6] passed; build-classic-hosts.py --check OK; scan-secrets no leaks found; e2e tests passed (skipped=1); smoke-install passed; test-observability passed, source tree clean. Not run: go test, go vet, check-public.py — no Go or public-surface changes in this PR.
Safety
All four safety boxes checked: no credentials/internal endpoints/production data/organization-specific infrastructure; preserved user files and unrelated changes; updated generated Classic host bundles when applicable; documented third-party content and licensing changes when applicable.