rsicarelli/fakt / .claude/skills/skill-creator
skill-creator
VerifiedCreates new Claude Code Skills following project patterns with trigger-rich descriptions and progressive disclosure. Use when creating new skills, converting commands to skills, scaffolding skill structure, or updating an existing skill's structure. Make sure to use this skill whenever a new skill needs to be created — it enforces the project's conventions for descriptions, body structure, allowed-tools, and progressive disclosure that are easy to get wrong.
Skill Creator
Scaffolds new Claude Code Skills with proper structure, trigger-rich descriptions, and progressive disclosure.
Instructions
1. Gather Requirements
Required:
- Skill name (kebab-case)
- Core purpose (what problem does it solve?)
- Trigger keywords (how will users invoke it?)
Optional:
- Required tools (Read, Bash, etc.)
- Supporting files needed
- Dependencies on other skills
If migrating from slash command:
cat .claude/commands/{command-name}.md
# Extract: allowed-tools, core logic, dependencies, arguments
2. Craft Description
Descriptions must be "pushy" — they should actively combat under-triggering by covering edge cases and using assertive language. The goal is to make sure the skill gets activated whenever it's relevant.
Template:
{What it does}. Use when {scenario 1}, {scenario 2}, {scenario 3}, or {scenario 4}. Make sure to use this skill whenever {broader activation context} — {reason why this matters}.
Rules:
- Max 1024 chars
- Must include "Use when" clause with 3+ specific scenarios
- Must include "Make sure to use this skill whenever" for broader activation
- Third person (never "I can" or "you can")
- Be specific, not vague
- No quoted keyword lists — trust semantic matching
- Explain why the skill matters, not just what it does
Example:
# Bad: Validates Kotlin APIs
# Mediocre: Validates Kotlin compiler API usage against source code. Use when validating API calls or checking compatibility.
# Good: Validates Kotlin compiler API usage against source code, checking deprecations and compatibility. Use when validating API calls, checking compiler API compatibility, verifying API hasn't been deprecated, or debugging API issues. Make sure to use this skill whenever compiler plugin code references Kotlin internal APIs — these APIs change frequently between versions and silent breakage is common.
3. Design Structure
Simple (no supporting files):
skill-name/
└── SKILL.md (<300 lines)
Medium (some references):
skill-name/
├── SKILL.md # <500 lines
└── resources/
└── reference.md # On-demand
Complex (scripts + docs):
skill-name/
├── SKILL.md
├── scripts/
└── resources/
Rule: If SKILL.md would exceed 500 lines, extract details to resources/.
Writing Style: Explain the "Why"
Instructions should explain why each step exists, not just what to do. This helps the agent make better decisions when encountering edge cases.
# Bad — just commands:
### 3. Run tests
Run `make test-sample`.
# Good — explains why:
### 3. Run tests
Run `make test-sample` to verify generated code compiles against the published plugin.
This catches issues that unit tests miss because it exercises the full compilation pipeline.
Keep instructions lean and imperative. Add context lines only where the "why" isn't obvious.
4. Generate SKILL.md
---
name: {skill-name}
description: {description from step 2}
allowed-tools: {minimal set — only what's needed}
---
# {Skill Title}
{One-line purpose.}
## Instructions
### 1. {First Step}
{Details}
### 2. {Second Step}
{Details}
...
## Supporting Files
{List resources/ if any}
## Related Skills
{Skills this composes with}
Allowed-tools guidelines:
- Analysis:
Read, Grep, Glob - Execution:
Read, Bash, TaskCreate, TaskUpdate - Generation:
Read, Write, Grep, Glob - Complex:
Read, Write, Edit, Bash, Grep, Glob, TaskCreate, TaskUpdate
5. Create Files
SKILL_NAME="{skill-name}"
mkdir -p ".claude/skills/${SKILL_NAME}/resources"
# Write SKILL.md and any supporting files
6. Verify
- SKILL.md has valid YAML frontmatter
- Description < 1024 chars
- Description includes "Use when" with 3+ scenarios
- Description includes "Make sure to use this skill whenever" (pushy tone)
- allowed-tools is minimal
- Instructions are numbered, clear, and explain "why" where non-obvious
- Supporting files documented with
## Supporting Filesif resources/ exists - SKILL.md body < 500 lines (extract to resources/ if over)
7. Summary
SKILL CREATED: {skill-name}
Location: .claude/skills/{skill-name}/
Files: SKILL.md ({lines} lines), resources/ ({count} files)
Description: {description}
Tools: {allowed-tools}
Test by asking naturally: "{example prompt that should trigger this skill}"
Related Skills
codegen,feature-option— good examples of existing skill patterns
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 467k | Updated | Claude CodeCodex | Verified | Install |
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 467k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse 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. | Stars: 319k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 319k | Updated | Claude CodeCodex | Verified | Install |