agentleFS
calebnewtonusccalebnewtonusc/Chewbacca/skills/skill-forge/SKILL.md

skill-forge

Skill8 starsChanged 3 months agoMITView on GitHub

Write a new skill, or fix one that is not being obeyed. Use when the user asks for a new skill, says a skill is being ignored or does the wrong thing, wants to turn a repeated task into something reusable, or asks to install a skill someone else wrote. Also use before writing any SKILL.md, because the authoring order matters more than the file.

49 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use sigma, agent-forge and 2 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.
  • sanyuan0704Skill

    sigma

    Personalized 1-on-1 AI tutor using Bloom's 2-Sigma mastery learning. Guides users through any topic with Socratic questioning, adaptive pacing, and rich visual output (HTML dashboards, Excalidraw concept maps, generated images). Use when user wants to learn something, study a topic, understand a concept, requests tutoring, says 'teach me', 'I want to learn', 'explain X to me step by step', 'help me understand', or invokes /sigma. Triggers on: learn, study, teach, tutor, understand, master, explain step by step.

    Also in 3 projects that use this

  • xyreinsurance119Skill

    agent-forge

    Создание, улучшение и аудит скиллов и агентов OpenClaw. Три режима: создание скилла (11 шагов), создание агента (9 шагов, с памятью и автоулучшением), улучшение существующего (5 шагов). Triggers: 'создай скилл', 'новый скилл', 'создай агента', 'новый агент', 'улучши скилл', 'skill creator', 'agent creator', 'скиллмейкер'.

    Also in 3 projects that use this

  • localSummerSkill

    skill-review

    回顾当前对话,整理 skill 使用情况、代码质量反馈和迭代建议,生成待确认的更新清单

    Also in 3 projects that use this

  • GoogleSkill

    code-review

    Adversarial systems, concurrency, memory-model, and architecture code review protocol. Use this skill whenever asked to review code, PRs, diffs, or branches in Filament.

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

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

  • 0

    Common

    Most of the 42 edited versions we read keep these.

  • 3

    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:

    Name the pain, then the problem, then the prompt: If you cannot say what specifically goes wrong today, you are building for a problem …

    line 13
  2. Distinctive:

    Run it by hand as a plain prompt: Not once. Until the output is right and you know why it is right.

    line 16
  3. Distinctive:

    If skills are listed anywhere, generate the list from the frontmatter and add a check that fails when it drifts.

    line 78

“Common” is measured: kept by at least 6 in 10 of the 42 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

skill-forge

Most skills fail the same way: somebody wrote the file before they had run the task. The file then encodes a guess, and the guess is obeyed.

The order is the whole method

  1. Name the pain, then the problem, then the prompt. If you cannot say what specifically goes wrong today, you are building for a problem nobody has. Stop here more often than feels comfortable.
  2. Run it by hand as a plain prompt. Not once. Until the output is right and you know why it is right.
  3. Write the playbook. Five parts, all of them: the rules, the steps, the standards for good output, worked examples, and the never-dos. This is prose, not a file format.
  4. Only now write SKILL.md, from the playbook.
  5. Test it against the cases that broke in step 2.

Skipping from 1 to 4 is the failure this skill exists to prevent.

Rules for the file itself

One task per skill. The test: would you ever run this task on its own? If no, it is a step, not a skill. Skills that do five things or chain several events do not get obeyed. Chain narrow skills instead.

Open with one sentence saying what this is, above all the detail. Detail placed above purpose outvotes the purpose.

Length is a constraint, not a budget. A long skill stops being followed. Instructions belong in SKILL.md. Facts about a project, a person or a corpus belong in a referenced file the skill points at.

Write non-negotiables as prohibitions, not as tone. "Never invent a client name" is enforceable. "Be accurate" is not. Anything that goes out under a real name needs its prohibitions written as hard lines.

Include the negative examples. What a wrong answer looks like teaches more than another right one.

A human-in-the-loop step is a feature. A skill that interviews you before producing is often better than one that one-shots. Do not optimize that away.

When a chained step keeps returning the same answer

Give it its own turn rather than more instructions. A step folded into a larger skill collapses toward its own first answer. Splitting it out fixes what more prompting does not.

The self-score gate

For anything with a quality bar, make the skill score its own output against its own rules and refuse to deliver under a threshold. It converts standards you wrote down into standards that get checked.

Never install a found skill unopened

A skill from a marketplace or a repo is inspiration, not a dependency. Read what it actually does first: it runs with your permissions, your files and your credentials. Prefer add-skill.sh, which clones rather than vendors and prints the declared license before installing, because a badge saying Apache 2.0 can sit on top of a proprietary or AGPL skill that relicenses a project by contagion.

Naming and versioning

Kebab-case directory matching the skill name. When a skill is versioned, put the version at the front of the title, because names truncate from the right and the one question worth answering at a glance is whether this is current.

Registry drift is a test, not a paragraph

If skills are listed anywhere, generate the list from the frontmatter and add a check that fails when it drifts. A hand-maintained index rots, and the rot is silent.

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.