skills/skill/SKILL.md
Create new Claude Code skills following project patterns and best practices. Use when building new skills, extracting reusable capabilities, or converting commands to skills.
npx skillsauth add srnnkls/tropos skillInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Create well-structured skills using progressive disclosure and project conventions.
Reference: best-practices.md for comprehensive guidance, reference.md for project patterns and frontmatter specs.
Gather concrete examples of how the skill will be used:
Skip this step only when usage patterns are already clearly understood.
Analyze each use case to identify reusable resources:
| Resource Type | When to Use | Example |
|---------------|-------------|---------|
| scripts/ | Same code rewritten repeatedly | rotate_pdf.py |
| reference/ | Domain knowledge Claude needs | schema.md, api.md |
| assets/ | Files used in output | template.html, logo.png |
| templates/ | Document structure patterns | report.md |
Required fields:
name: skill-name # Lowercase, hyphens, max 64 chars
description: | # Max 1024 chars
[What it does]. Use when [context].
Optional fields:
| Field | Purpose | Example |
|-------|---------|---------|
| context | Run in forked subagent | context: fork |
| agent | Specify agent type | agent: haiku |
| user-invocable | Hide from slash menu | user-invocable: false |
| allowed-tools | Restrict available tools | See reference.md |
| hooks | Lifecycle hooks (PreToolUse, PostToolUse, Stop) | See reference.md |
Naming pattern: <namespace>[-<subnamespace>]-<action>
dispatch, scope, git worktreeDescription format: Third person, what + when.
mkdir -p .claude/skills/{skill-name}
Standard structure:
.claude/skills/{skill-name}/
├── SKILL.md # Main instructions (<500 lines)
├── templates/ # Document templates (.md)
├── scripts/ # Executable code (.sh, .py)
└── reference/ # Extended documentation
Write SKILL.md:
Test with real tasks:
Match specificity to task fragility:
High freedom - Multiple approaches valid, context-dependent:
## Code review
1. Analyze structure and organization
2. Check for bugs and edge cases
3. Suggest improvements
Low freedom - Operations fragile, consistency critical:
## Database migration
Run exactly: `python scripts/migrate.py --verify --backup`
Do not modify flags.
| Type | Characteristics | Examples |
|------|-----------------|----------|
| Operational | Multi-step workflow, state changes, document templates | scope |
| Generation | Transform input → structured output, format templates | docs-implement |
| Guidance | Imperative instructions, code patterns | implement, loqui |
<namespace>[-<subnamespace>]-<action>tools
External code-review harness (`peer` zsh tool): canonical model registry, idle-stall watchdog, and self-parallelising fan-out to codex/gemini. Use when dispatching external (non-Claude) reviewers from the review or implement pipelines — call `peer run`/`peer <harness>` instead of `codex exec`/`gcloud`+Vertex directly.
testing
Unified validation dispatcher. Auto-detects validation type from argument or presents selection menu. Routes to test, implement (verify), or hooks-test.
development
Test-driven development methodology (RED-GREEN-REFACTOR). Use when implementing features, fixing bugs, or changing behavior - write failing test first, then minimal code to pass.
tools
Unified scope lifecycle. Auto-detects operation from argument or presents selection menu. Routes to create, review, update, done, or list.