agentleFS
s977043s977043/river-review/.claude/skills/skill-creator/SKILL.md

skill-creator

Skill7 starsChanged 5 months agoMITView on GitHub

Design a new Claude Code skill from a concrete use case and produce a repo-ready skill package. Use when the user asks to create a new skill, define a skill's responsibility, draft SKILL.md, choose frontmatter, design supporting files, or prepare eval criteria for a new skill. Also trigger on "スキルを作りたい", "スキルを作って", "スキルを追加して", "新しいスキル", "SKILL.md生成".

189 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use taskflow-inbox-triage, mermaid-diagram-generator 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.
  • AtomicBot-aiSkill

    taskflow-inbox-triage

    Also in 6 projects that use this

  • shalombSkill

    mermaid-diagram-generator

    Guide on creating and editing Mermaid charts safely. Use when asked to draw a diagram, flowchart, or architecture chart. Triggers on: 'mermaid', 'diagram', 'flowchart', 'visualize'.

    Also in 5 projects that use this

  • naman10parikhSkill

    ecosystem-reference

    Quick-access directory of Claude Code ecosystem repos, skills, plugins, MCPs, and patterns. Use when improving the harness.

    Also in 5 projects that use this

  • naman10parikhSkill

    fractal-delegation

    Also in 5 projects that use this

  • Liang-MHSkill

    stock_analysis

    获取 A 股、基金和港股的实时行情和技术分析数据

    Also in 5 projects that use this

  • nicepkgSkill

    community-led-growth

    Expert in community-led growth (CLG) - leveraging user communities to drive acquisition, retention, and expansion. Covers building developer communities, user groups, ambassador programs, and turning customers into advocates. Knows the difference between community as a feature vs community as a growth engine, and how to measure community ROI. Use when "community-led, CLG, ambassador program, champion program, user community, developer community, word of mouth, user group, " mentioned.

    Also in 5 projects that use this

  • +6 more from naman10parikh

The file

Show its rules and text

We split this skill into the rules it gives an agent

24 rules. 0 are common, 17 are distinctive, and 7 are very specific to this project.

  • 0

    Common

    Most of the 185 edited versions we read keep these.

  • 17

    Distinctive

    This author's own choices. Worth a look.

  • 7

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Create a new skill only after the use case is concrete enough.

    line 15
  2. Very specific:

    Write output files only to .claude/skills/ unless the user explicitly specifies a different path.

    line 29
  3. Distinctive:

    Do not modify existing skills without user confirmation.

    line 30
  4. Distinctive:

    Do not create a broad or fuzzy skill.

    line 36
  5. Distinctive:

    If the requested scope is too broad, split it into multiple skills.

    line 44
  6. Distinctive:

    If any item is unclear:

    line 59
  7. Distinctive:

    ask the minimum focused questions, or

    line 61
  8. Distinctive:

    Do not draft the full skill until the gate passes.

    line 64
  9. Distinctive:

    add secondary patterns only when one pattern is insufficient

    line 82
  10. Distinctive:

    Do not continue until the pattern choice is explicit.

    line 85
  11. Very specific:

    Keep the main file compact.

    line 131
  12. Very specific:

    Use ${CLAUDE_SKILL_ROOT}/assets/basic-skill-template.md as the starting skeleton.

    line 135
  13. Distinctive:

    Create an initial eval plan with:

    line 141
  14. Very specific:

    Use ${CLAUDE_SKILL_ROOT}/assets/eval-rubric-template.md as the rubric structure.

    line 155
  15. Distinctive:

    Check the draft against:

    line 159
  16. Distinctive:

    After the skill files are written:

    line 168
  17. Very specific:

    verify the directory name matches the name field in frontmatter

    line 170
  18. Very specific:

    verify description is under 1024 characters and includes trigger keywords

    line 171
  19. Very specific:

    verify all ${CLAUDE_SKILL_ROOT}/ references point to existing files

    line 174
  20. Distinctive:

    check the repository CLAUDE.md or AGENTS.md for additional registration steps

    line 175
  21. Distinctive:

    if the repository uses a skills index or registry, update it

    line 176
  22. Distinctive:

    create a "do everything" skill

    line 200
  23. Distinctive:

    write a vague description

    line 202
  24. Distinctive:

    If the request is too broad:

    line 211

“Common” is measured: kept by at least 6 in 10 of the 185 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

Pattern declaration

Primary pattern: Pipeline Secondary patterns: Inversion Why: skill creation requires strict phase order (requirements → pattern selection → structure → review) with an upfront question-gathering gate.

Purpose

Create a new skill only after the use case is concrete enough.

Your job is to:

  1. identify the smallest useful responsibility
  2. select the right design pattern
  3. define when the skill should trigger
  4. choose safe frontmatter defaults
  5. draft a minimal but strong SKILL.md
  6. propose supporting files only when they add clear value
  7. define an initial eval set

Safety constraints

Write output files only to .claude/skills/ unless the user explicitly specifies a different path. Do not modify existing skills without user confirmation.

Note: this skill does not set disable-model-invocation: true because file creation is its primary purpose, not an incidental side effect. The settings.json deny list and user confirmation prompts provide sufficient guardrails.

Core rule

Do not create a broad or fuzzy skill.

Prefer:

  • one responsibility
  • one primary workflow
  • one clear success condition

If the requested scope is too broad, split it into multiple skills.

Phase 0: Gate

Before drafting, confirm:

  • target outcome
  • primary user intent
  • trigger phrases
  • audience
  • allowed tools needed
  • whether the skill should be auto-invoked or manual-only
  • source material available
  • definition of done

If any item is unclear:

  • ask the minimum focused questions, or
  • inspect available files first if the answer can be found without asking

Do not draft the full skill until the gate passes.

Phase 0.5: Pattern selection

Choose the primary design pattern before drafting the skill.

Available patterns:

  • Tool Wrapper — the skill mainly injects domain, library, or framework knowledge on demand
  • Generator — the skill mainly produces structured output from a reusable template
  • Reviewer — the skill mainly evaluates output against a checklist or rubric
  • Inversion — the skill must gather missing information before acting
  • Pipeline — the skill must enforce a strict multi-step workflow with checkpoints

Selection rules:

  1. identify the primary value source or failure mode the skill addresses
  2. map it to one primary pattern
  3. add secondary patterns only when one pattern is insufficient
  4. each secondary pattern must appear in a named phase with a measurable role

Do not continue until the pattern choice is explicit.

Phase 1: Skill boundary design

Produce a short working summary with:

  • skill name candidate
  • single responsibility
  • in-scope tasks
  • out-of-scope tasks
  • likely trigger language
  • risk level

Reject designs that combine unrelated workflows.

Phase 2: Frontmatter design

Choose frontmatter intentionally.

Defaults:

  • include name
  • include description
  • omit extra fields unless clearly needed

Use:

  • disable-model-invocation: true for side-effecting, sensitive, or timing-dependent workflows
  • allowed-tools only when safe pre-approval is clearly useful
  • model only when there is a concrete reason to override the default

For every chosen field, state why it is needed.

Phase 3: Structure design

Draft the skill with this order:

  1. pattern declaration
  2. purpose
  3. gate
  4. workflow phases
  5. output contract
  6. review step
  7. failure handling
  8. additional resources

Keep the main file compact. Move heavy references, examples, and templates into separate files. For River Review Agent Skills, keep activation, responsibility, routing, hard guards, and the output contract in SKILL.md; move optional examples, troubleshooting, rationale, and provider-specific detail to references/. Treat the repository's 12 KiB / 250-line context-health threshold as an advisory review signal, never as a reason to remove safety or verification.

Use ${CLAUDE_SKILL_ROOT}/assets/basic-skill-template.md as the starting skeleton.

The generated SKILL.md must include a Pattern declaration section at the top with primary pattern, secondary patterns, and Why fields matching the choice from Phase 0.5.

Phase 4: Eval design

Create an initial eval plan with:

  • 10 to 20 realistic test prompts
  • 3 to 6 pass/fail criteria
  • likely failure modes
  • what should be checked by code vs by reviewer

Include at least:

  • one trigger test
  • one under-specified request
  • one boundary case
  • one anti-pattern case

Use ${CLAUDE_SKILL_ROOT}/assets/eval-rubric-template.md as the rubric structure.

Phase 5: Review

Check the draft against:

  • ${CLAUDE_SKILL_ROOT}/references/review-default.md
  • ${CLAUDE_SKILL_ROOT}/references/design-principles.md

Fix any issue before finalizing.

Phase 6: Registration

After the skill files are written:

  1. verify the directory name matches the name field in frontmatter
  2. verify description is under 1024 characters and includes trigger keywords
  3. measure the SKILL.md entrypoint size and review whether optional knowledge should move to supporting files
  4. for River Review Agent Skills, run npm run agent-skills:validate and inspect any context-health advisory; exceeding 12 KiB / 250 lines is warning-only and requires judgment, not blind trimming
  5. verify all ${CLAUDE_SKILL_ROOT}/ references point to existing files
  6. check the repository CLAUDE.md or AGENTS.md for additional registration steps
  7. if the repository uses a skills index or registry, update it

Output format

Return exactly these sections:

Skill summary

Pattern selection

Recommended directory structure

SKILL.md draft

Supporting files to add

Eval plan

Known risks

Anti-patterns

Do not:

  • create a "do everything" skill
  • invent company-specific rules
  • write a vague description
  • skip the gate
  • skip pattern selection
  • omit eval planning
  • hide trade-offs
  • stack multiple patterns without justifying each one

Failure handling

If the request is too broad:

  • propose 2 to 4 smaller skills
  • explain the split by responsibility
  • recommend which one to build first

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.