agentleFS
markoblogomarkoblogo/abvx-agent-skills/skills/phase-spec-execution/SKILL.md

phase-spec-execution

Skill16 starsChanged 5 months agoMITView on GitHub

Execute large tasks through explicit phases with acceptance criteria, required commands, and state updates. Use when work is too large for one implementation pass but too important to leave as a vague checklist. Keep phases falsifiable, end-to-end, and small enough that each one can be verified honestly before moving on.

The file

We split this skill into the rules it gives an agent

9 rules. About 0 are common, 9 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈9

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Update state after each phase so the next pass does not rediscover context.

    line 26
  2. Distinctive:

    Do not create more phases than the task needs.

    line 27
  3. Distinctive:

    Write or state one clear objective per phase.

    line 45
  4. Distinctive:

    Before implementing a phase, restate its acceptance criteria.

    line 46
  5. Distinctive:

    Run the phase verification commands.

    line 48
  6. Distinctive:

    Only then advance to the next phase.

    line 50
  7. Distinctive:

    Do not create "phase 1: backend, phase 2: frontend, phase 3: tests" by default.

    line 64
  8. Distinctive:

    Do not advance on "looks good" when acceptance criteria are falsifiable.

    line 65
  9. Distinctive:

    Do not let plan files become larger than the implementation itself.

    line 66

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “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 Spec Execution

Turn a large task into a sequence of falsifiable execution slices instead of one broad blob of work.

Use When

  • the task spans multiple files, layers, or user-visible milestones;
  • there is real sequencing or dependency ordering;
  • you need clearer state than "still working";
  • the user wants autonomy without losing verification discipline.

Core Rules

  • Slice vertically where possible. Each phase should move observable behavior.
  • Give every phase explicit acceptance criteria.
  • Attach the narrowest mandatory commands that can prove the phase.
  • Update state after each phase so the next pass does not rediscover context.
  • Do not create more phases than the task needs.

Phase Template

For each phase, define:

  • objective;
  • dependencies or prerequisites;
  • files or surfaces likely to change;
  • acceptance criteria;
  • mandatory verification commands;
  • known risks;
  • handoff note if the phase cannot complete.

Workflow

  1. Identify the end state the user actually wants.
  2. Break the work into the smallest useful number of ordered phases.
  3. Write or state one clear objective per phase.
  4. Before implementing a phase, restate its acceptance criteria.
  5. Execute only the current phase.
  6. Run the phase verification commands.
  7. Record what changed, what passed, and what remains.
  8. Only then advance to the next phase.

State Discipline

Track at least:

  • current phase;
  • completed phases;
  • failed checks and why;
  • assumptions that changed;
  • open risks for the next phase.

Anti-Patterns

  • Do not create "phase 1: backend, phase 2: frontend, phase 3: tests" by default.
  • Do not advance on "looks good" when acceptance criteria are falsifiable.
  • Do not let plan files become larger than the implementation itself.

Final Report

Report completed phases, skipped phases, criteria met, criteria deferred, and any remaining integration risk.

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.