agentleFS
CarriedWorldUniverseCarriedWorldUniverse/nexus/.agents/skills/writing-skills/SKILL.md

writing-skills

Skill0 starsChanged 36 days agoApache-2.0View on GitHub

Use when writing, editing, reviewing, or pruning a skill (SKILL.md) — the craft reference for making skills predictable: invocation choice, description writing, information hierarchy, leading words, and the named failure modes (premature completion, duplication, sediment). Also use for the periodic pruning pass over ~/.claude/skills/.

191 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use chinese-code-review, chinese-commit-conventions and 10 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.
  • Leodorareluctant259Skill

    chinese-code-review

    中文代码审查规范——在保持专业严谨的同时,用符合国内团队文化的方式给出有效反馈

    Also in 8 projects that use this

  • Leodorareluctant259Skill

    chinese-commit-conventions

    中文 Git 提交规范 — 适配国内团队的 commit message 规范和 changelog 自动化

    Also in 8 projects that use this

  • obraSkill

    diagnosing-superpowers

    Use when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.

    Also in 20 projects that use this

  • kntismSkill

    condition-based-waiting

    Use when tests have race conditions, timing dependencies, or inconsistent pass/fail behavior - replaces arbitrary timeouts with condition polling to wait for actual state changes, eliminating flaky tests from timing guesses

    Also in 3 projects that use this

  • aaddrickSkill

    writing-agents

    Use when creating new agents, editing existing agents, or defining specialized subagent roles for the Task tool

    Also in 3 projects that use this

  • pedronauckSkill

    lesson-learned

    Analyze recent code changes via git history and extract software engineering lessons. Use when the user asks 'what is the lesson here?', 'what can I learn from this?', 'engineering takeaway', 'what did I just learn?', 'reflect on this code', or wants to extract principles from recent work.

    Also in 3 projects that use this

  • +3 more from Leodorareluctant259

The file

We split this skill into the rules it gives an agent

1 rule. 0 are common, 1 is distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 181 edited versions we read keep these.

  • 1

    Distinctive

    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:

    Put the skill's leading word first.

    line 23

“Common” is measured: kept by at least 6 in 10 of the 181 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

Writing skills well

A skill exists to wrangle determinism out of a stochastic system. The root virtue is predictability — the agent taking the same process every run (not producing the same output). Every rule below serves it.

(Adapted from mattpocock/skills writing-great-skills, fitted to our conventions. Our prior rule stands: skills = procedures + the tools to run them; facts/state/decisions → memory; needed-every-turn → CLAUDE.md. A skill that is mostly facts is a memory file wearing the wrong hat.)

Invocation: two costs, pick one

  • Model-invoked (has a rich description): the agent can fire it autonomously and other skills can reach it. Costs context load — the description sits in the window every turn, forever.
  • User-invoked (disable-model-invocation: true, human-facing one-line description): zero context load, but costs cognitive load — the operator is the index that must remember it exists.

Pick model-invocation only when the agent must reach it on its own or another skill must reach it. Our per-role SkillAllowlist work sharpens the stakes: every model-invoked description a role can see is context that role pays for on every turn — scope tightly.

The description

Two jobs: state what the skill is, and list the branches that trigger it. It earns the hardest pruning in the file:

  • Put the skill's leading word first.
  • One trigger per branch — synonyms restating one branch are duplication; collapse them.
  • Cut identity that's already in the body; keep triggers + any "when another skill needs…" reach clause.

Information hierarchy

Two content types — steps (ordered actions) and reference (rules/definitions consulted on demand) — mix freely. Place each piece on the ladder by how immediately the agent needs it:

  1. In-skill step — ends on a completion criterion that is checkable (done vs not-done decidable) and, where it matters, exhaustive ("every modified file accounted for", not "produce a list"). A vague criterion invites premature completion. This is the same rule as the broker's observable-acceptance-criteria doctrine — one discipline, two homes.
  2. In-skill reference — a flat peer-set of rules on one rung is fine, not a smell.
  3. External reference — pushed to a sibling file behind a context pointer, loaded only when it fires. The pointer's wording decides whether the agent reaches it.

Progressive disclosure = moving material down the ladder so the top stays legible. The branching test: inline what every branch needs; push behind a pointer what only some branches reach. Co-location: keep a concept's definition, rules, and caveats under one heading — reading one part brings its neighbours.

Leading words

A leading word is a compact concept already in the model's pretraining that anchors a region of behaviour in one token (tracer bullet, fog of war, red, seal, fold). It serves predictability twice: in the body it anchors execution; in the description it anchors invocation. Hunt for restatements begging to collapse: "fast, deterministic, low-overhead" → tight; "a loop you believe in" → red. Fewer tokens AND a sharper hook. (Our own examples already work this way: cairn's express/fold/seal, the gate work's judge-the-diff, the cognitive-core's pool/belief.)

When to split

Each cut spends one of the two loads — split only when the cut earns it:

  • By invocation: a distinct leading word should trigger independently, or another skill must reach it. Pays context load.
  • By sequence: the steps still ahead tempt the agent to rush the current one — hide the post-completion steps. Use only when the completion criterion is irreducibly fuzzy AND the rush is observed; sharpening the criterion is cheaper and comes first.

Failure modes (diagnose by name)

  • Premature completion — ending a step before it's genuinely done. Defence in order: sharpen the completion criterion (cheap, local); only then split the sequence.
  • Duplication — one meaning in two places. Costs maintenance + tokens, and inflates the meaning's rank on the ladder. Keep a single source of truth per meaning.
  • Sediment — stale layers that settle because adding feels safe and removing feels risky. The default fate of any skill without a pruning discipline. Our verify-the-deployed-artifact lesson generalizes: a skill's text drifts from practice exactly like a manifest drifts from a deploy.

The pruning pass (run this over a skill, or the whole library)

For each skill:

  1. Hat check — is any section facts/state/decisions? Move it to memory; leave a pointer if a branch needs it.
  2. Description audit — one trigger per branch, leading word front-loaded, no body-identity restated.
  3. No-op hunt — test each sentence in isolation; a sentence that changes no behaviour is deleted whole, not trimmed.
  4. Duplication sweep — every meaning has one home; collapse restatements into the leading word.
  5. Ladder check — anything inline that only one branch reaches → push behind a pointer; anything behind a pointer that every run needs → pull up.
  6. Criterion check — every step ends on a checkable, exhaustive-where-it-matters completion criterion.
  7. Sediment date — if a line describes a practice we no longer follow, delete it now; "might need it" is what memory and git history are for.

Completion criterion for the pass itself: every skill in scope has had all seven checks applied and each resulting edit either made or explicitly declined with a reason.

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.