agentleFS
forsonnyforsonny/DG-Anvil/skills/authoring-skills/SKILL.md

authoring-skills

Skill0 starsChanged 7 months agoMITView on GitHub

Meta-skill governing every change to any skill file. Requires a RED-then-GREEN subagent pressure transcript before a skill change is accepted.

9 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

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

  • 0

    Common

    Most of the 8 edited versions we read keep these.

  • 3

    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. Very specific:

    Create or update the paired pressure test file at tests/pressure/<skill>.pressure.js.

    line 17
  2. Distinctive:

    Run the test with the skill absent.

    line 19
  3. Very specific:

    Write or amend skills/<name>/SKILL.md.

    line 20
  4. Distinctive:

    Run the test with the skill present.

    line 21
  5. Distinctive:

    If any of these conditions obtain, the skill change is rejected:

    line 35
  6. Very specific:

    No paired tests/pressure/<skill>.pressure.js file in the same change set.

    line 37

“Common” is measured: kept by at least 6 in 10 of the 8 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

Overview

Authoring a skill is test-driven. A skill change is not merged until a paired pressure run fails without the change and passes with it. The pressure test cites a failure-taxonomy row from docs/failure-taxonomy.md. The skill description is a trigger for activation, not a summary of the body.

When to Use

Whenever any skills/*/SKILL.md file is created or modified. Whenever a pressure test under tests/pressure/ is added, removed, or rewritten. Whenever the canonical section order, skill inventory, or skill-retirement policy is changed.

Process

  1. Identify the failure-taxonomy row the new or amended skill defeats. Write the row number in the pressure test as a single-line comment at the top of the file.
  2. Create or update the paired pressure test file at tests/pressure/<skill>.pressure.js. Import runPressure from tests/pressure/harness.js.
  3. Define a scenario: a prompt, a minimal repository state, and the assertions that distinguish a skill-absent outcome from a skill-present outcome.
  4. Run the test with the skill absent; confirm the RED outcome (expected failure mode reproduces).
  5. Write or amend skills/<name>/SKILL.md. Keep the six canonical H2 sections in order: Overview, When to Use, Process, Rationalizations, Red Flags, Verification.
  6. Run the test with the skill present; confirm the GREEN outcome (skill prevents the failure).
  7. Review the description: it must trigger activation, never summarise the body.
  8. Adding an eighth skill? Retire one. The skill inventory is capped at seven; each addition requires an explicit retirement in the same change.

Rationalizations

Reject the following shortcuts:

  • "The description is a trigger, not a summary." Authors violate this by writing a description that restates the Overview. The skill will then never be activated at the right time (row 18: description-field shortcut).
  • "One more skill is harmless." Skill proliferation erodes the triggering signal for every existing skill (row 29: skill-proliferation drag).
  • "The pressure test can come later." Without a RED-then-GREEN transcript, there is no evidence the skill defeats anything. Defer the merge, not the test (row 18 again).

Red Flags

If any of these conditions obtain, the skill change is rejected:

  • No paired tests/pressure/<skill>.pressure.js file in the same change set.
  • The pressure test does not cite a failure-taxonomy row by number.
  • The skill description summarises the body instead of naming an activation trigger (row 18).
  • The change adds a new skill without retiring an existing one (row 29).

Verification

A reviewer accepts a skill change only after confirming, by file inspection:

  1. The pressure test file exists and cites a failure-taxonomy row.
  2. Both the skill-absent and skill-present runs are recorded with their outcomes.
  3. The skill description is a trigger phrase, not a summary of the body.
  4. The six canonical sections are present in order with non-empty bodies for any skill other than the bootstrap loader.

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.