agentleFS
yokeloopyokeloop/yoke/.claude/skills/yoke-validate/SKILL.md

yoke-validate

Skill1 starsChanged 4 months agoMITView on GitHub

Validates SKILL.md files changed in the current branch against elements-of-style (Strunk) for prose and plugin-dev's skill-development conventions for structure, applies safe fixes automatically, and reports the rest. Activates when the user writes "validate", "validate skill", "validate skills", "validate changed skills", "lint skills", "валидайт", "валидайт скилы", "прогони валидацию", "проверь скилы".

The file

We split this skill into the rules it gives an agent

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

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 5

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Run two lenses on every SKILL.md changed in this branch and auto-fix the safe findings.

    line 14
  2. Distinctive:

    If either lens is not installed, stop and name the missing dependency.

    line 23
  3. Very specific:

    If the result is empty (the current branch is main, or no SKILL.md differs from main), report "nothing to validate" and stop.

    line 33
  4. Very specific:

    Every referenced file (${CLAUDE_PLUGIN_ROOT}/..., references/..., etc.) resolves on disk.

    line 51
  5. Distinctive:

    Track: every applied fix and every unresolved finding for this file.

    line 52
  6. Distinctive:

    After processing every file, print one combined report:

    line 56
  7. Very specific:

    Never git add, git commit, or otherwise mutate git state.

    line 68
  8. Very specific:

    If elements-of-style:writing-clearly-and-concisely or plugin-dev:skill-development is missing, stop and report the missing dependency by …

    line 69

≈ “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

Validate

Run two lenses on every SKILL.md changed in this branch and auto-fix the safe findings.

Lenses

Both lens skills must be installed locally:

  • elements-of-style:writing-clearly-and-concisely — invoke via the Skill tool, then Read its elements-of-style.md reference. Drives prose tightening.
  • plugin-dev:skill-development — invoke via the Skill tool (or Read the cached SKILL.md directly if the loader misbehaves). Drives structural validation.

If either lens is not installed, stop and name the missing dependency. Do not inline the lens's rules.

Scope

Process only */SKILL.md paths returned by:

git diff --name-only main...HEAD | grep -E '/SKILL\.md$' | sort -u

If the result is empty (the current branch is main, or no SKILL.md differs from main), report "nothing to validate" and stop. Skip commands, agents, references, CLAUDE.md, and docs/.

Process per file

For each path:

  1. Read the file.
  2. Apply elements-of-style. For each Strunk violation, apply the fix via Edit when safe:
    • Drop needless words (Rule 13): that is, in order to → to, actually, really, weakening very.
    • Active voice (Rule 10): rewrite passive when the actor is clear.
    • Positive form (Rule 11): convert not X to a positive antonym where natural.
    • Concrete language (Rule 12): replace vague quantifiers (a lot of → substantial, several types of → several).
    • Record judgement-call rewrites that risk changing meaning as unresolved findings instead of applying them.
  3. Apply plugin-dev's skill-development checklist.
    • Frontmatter contains name and description. Missing fields → add (derive description from the body when obvious; otherwise record as unresolved).
    • description uses third person and lists concrete trigger phrases.
    • Body uses imperative/infinitive form throughout. Convert every you, your, you'll, you're to imperative via Edit. Apply across the whole body, not only the diff hunks against main.
    • Body word count ≤ 2000 (target) / ≤ 5000 (hard ceiling). Over the ceiling → record "split into references/" as unresolved.
    • Every referenced file (${CLAUDE_PLUGIN_ROOT}/..., references/..., etc.) resolves on disk. Broken references → record as unresolved.
  4. Track every applied fix and every unresolved finding for this file.

Report

After processing every file, print one combined report:

  • Files processed — count and paths.
  • Auto-applied fixes per file — one-line summary each.
  • Unresolved findings per file — judgement calls and structural issues that need human input.
  • Files with no findings → list as OK.

Rules

  • Process only */SKILL.md. Skip commands, agents, references, docs, and CLAUDE.md.
  • Auto-fix safe and mechanical findings; defer judgement calls to the report.
  • Apply second-person → imperative across the whole body of every changed file, not only the diff hunks. The repo accepts the resulting rewrite cost in exchange for strict plugin-dev conformance.
  • Never git add, git commit, or otherwise mutate git state. Leave the working tree dirty; the user commits through /yoke:gca.
  • If elements-of-style:writing-clearly-and-concisely or plugin-dev:skill-development is missing, stop and report the missing dependency by name.
  • Language: report in the conversation language; edits to SKILL.md stay in the file's original language (English, per yoke convention).

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.