Skip to content
0xlloydPublic

About

An AI skill that designs the process the best practitioners would use for your task, runs it with approval gates, and saves it so the next run starts from a proven flow.

Topics

Resources

Stars

0 stars

Watchers

0 watching

Forks

Latest commit

 

History

4 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 

Repository files navigation

Flow

Design the process the best practitioners would use. Run it. Keep it.

Flow is an AI skill for big tasks. Instead of pushing every task through the same fixed pipeline, it researches how experts actually do that kind of work, composes a task-specific workflow, runs it with approval gates, and saves the workflow to a library so the next similar task starts from a proven process.

A marketing piece needs references and a copy pass. A video needs a shot list and a review of takes. A business case needs criteria and costings. Flow works out which steps the task needs, and why.

How a run works

Step What happens
1. Frame the task Goal, audience, definition of done, constraints; played back for confirmation
2. Check the flow library A saved flow that fits is run as the baseline; related ones lend parts
3. Research how the work is best done Reframe the task as a top professional would, find the bar-setters, and research the stages and checks experts insist on
4. Compose the flow Start from a six-step spine (Understand, Research, Plan, Execute, Verify, Learn), add modules, give each step a reason, a tool and a "done when"
5. Present the flow A compact table for the user to approve, edit or reorder
6. Run the flow Step by step, with a plan gate and options for judgement calls
7. Verify Always runs: checks against done-when, the flow's checks and skill rules; opens what was made
8. Review the run After-action review: intended, happened, why, change or repeat
9. Save the flow One markdown file per task type in a library folder the user owns, merged with a dated changelog

Say "just run it" to skip the check-ins and stop only at the plan.

Built on

  • Situational method engineering: compose a method from stored pieces, chosen by the situation, and check it for consistency.
  • Toyota standard work and kaizen: the saved flow is the baseline; improvements edit the standard itself.
  • After-action review: four questions, causes not blame, keep what worked.
  • Checklist design: named pause points, short do-confirm lists of critical, easy-to-miss checks.

Try it

Use flow to build a launch campaign for this product brief.

/flow — I need a business case for replacing our reconciliation tool.

Use flow to make a 30-second product video. Just run it.

Install

The skill is the skills/flow folder. Keep references/ alongside SKILL.md.

  • Claude: zip the flow folder (not the whole repository) and upload it as a custom skill, following Anthropic's instructions.
  • Other hosts: use the host's skill-loading mechanism; the instructions are plain Markdown.

Flow works on its own. If the optional go-deeper, best-in-class or session-self-learning skills are installed, it uses them; otherwise it applies the same approach directly.

Contents

About

An AI skill that designs the process the best practitioners would use for your task, runs it with approval gates, and saves it so the next run starts from a proven flow.

Topics

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors