agentleFS
librosdeialibrosdeia/builder-method-claude-plugin/skills/iterate/SKILL.md

iterate

Skill0 starsChanged 5 days agoMITView on GitHub

Phase 5 of the Builder Method — check that the built product does its job, flow by flow, review Claude's decisions, sort feedback into fix now / next version / discard, and loop back to Build for fixes. Use when the first version is built and the user wants to test or improve it, or runs /builder:iterate.

13 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use test, writing-for-agents and 4 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • Microsoft AzureSkill

    test

    Also in 3 projects that use this

  • MicrosoftSkill

    writing-for-agents

    Writing documents for agents. Use when creating or editing skills, or modifying AGENTS.md or CLAUDE.md.

    Also in 3 projects that use this

  • AWSSkill

    retro

    Conduct a retrospective on a coding session.

    Also in 3 projects that use this

  • NVIDIASkill

    research

    Before designing a feature, research how comparable libraries solve the problem, check best practices, known pitfalls, and upstream changes. Use when starting a new feature, investigating an unfamiliar area, or the user asks to research something.

    Also in 4 projects that use this

  • SentrySkill

    review

    Fixture review skill.

    Also in 4 projects that use this

  • Hugging FaceSkill

    implement

    Make tests pass. Invoke after /write-tests produces failing tests.

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

5 rules. 0 are common, 2 are distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 13 edited versions we read keep these.

  • 2

    Distinctive

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If you have no shell tool, you can still record results the user tells you, but you can't launch the app.

    line 17
  2. Distinctive:

    If "fix now" has items:

    line 45
  3. Very specific:

    Add one task per fix to builder/04-plan.md, under a heading ## Cycle <new cycle> — fixes, in exactly the plan format, one line per field: …

    line 47
  4. Very specific:

    If "fix now" is empty, every flow is pass, and no decisions are pending:

    line 52
  5. Very specific:

    Update the progress page, and close by suggesting /builder:next.

    line 56

“Common” is measured: kept by at least 6 in 10 of the 13 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Phase 5 · Iterate

Goal: confirm that the product does its job, and collect feedback to improve it. Input: the built product, builder/02-prd.md, builder/decisions.md. Output: builder/05-validation.md, flow results in state.flows, and either a validated product or a new build cycle.

Read and follow:

  • ${CLAUDE_PLUGIN_ROOT}/skills/_shared/tone.md
  • ${CLAUDE_PLUGIN_ROOT}/skills/_shared/state-rules.md
  • ${CLAUDE_PLUGIN_ROOT}/skills/_shared/gates.md: apply the Iterate gate.

If you have no shell tool, you can still record results the user tells you, but you can't launch the app. Say so, and suggest Claude Code.

Set phases.iterate.status to in_progress and currentPhase to iterate.

Step 1 · validate

  1. Delegate to the qa subagent (Agent tool, subagent_type: "builder:qa"). It writes builder/05-validation.md from ${CLAUDE_PLUGIN_ROOT}/skills/iterate/templates/05-validation.md: one ## F<n> — <name> section per flow, with a checkbox per acceptance criterion. If the subagent tool isn't available, write it yourself following ${CLAUDE_PLUGIN_ROOT}/agents/qa.md.
  2. Unless the user already tested it, launch the app (the dev script), tell them how to open it, and walk them through one flow at a time: "Try this… Did you see that?"
  3. For each flow, record the outcome:
    • in the file: tick the boxes, set Result: pass or Result: fail, and add notes in the user's words;
    • in the state: set the flow's validation to pass or fail.
  4. Tell them which decisions you took on your own that deserve a look, and what to focus on before approving.

Step 2 · decisions

Go through every Status: pending entry in builder/decisions.md, in plain words. For each one, the user keeps it (Status: kept) or changes it (Status: changed, plus a line saying what to change; that change becomes a "fix now" item). Update state.decisions.pending to the number still pending (normally 0).

Step 3 · triage

Collect every failed flow, requested change and new idea in the "Triage" table of 05-validation.md, each marked:

  • fix now: needed for this version to work as the PRD says (failed flows usually go here);
  • next version: worth doing later. Add it to the PRD's "Non-goals" under "Next version";
  • discard.

Suggest a classification. The user decides.

Step 4 · loop

If "fix now" has items:

  1. Increase state.cycle by 1.
  2. Add one task per fix to builder/04-plan.md, under a heading ## Cycle <new cycle> — fixes, in exactly the plan format, one line per field: ### T<next n> — <title>, then Covers: <flows>, Check: …, Model: …, and Cycle: <new cycle>. Add them to state.tasks with cycle: <new cycle>.
  3. Set Build back to in_progress, with its build step pending, and set currentPhase to build. Leave Iterate in_progress.
  4. Prepare Iterate for re-checking the fixes: set every step in phases.iterate.steps back to pending, and set every flow whose validation is fail (and any flow a fix touches) back to pending. The next validation pass rewrites 05-validation.md for the new cycle, so the "fix now" items are not planned twice.
  5. Tell the user what will be fixed, and continue with /builder:build, unless they asked you to stop after planning.

If "fix now" is empty, every flow is pass, and no decisions are pending:

  • set review: reviewed in 05-validation.md, and mark Iterate and its steps done (loop is skipped);
  • set currentPhase to ship.

Update the progress page, and close by suggesting /builder:next.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.