Skip to content

SKILLS-CATALOG.md

Latest commit

 

History

History
437 lines (300 loc) · 14.8 KB

SKILLS-CATALOG.md

File metadata and controls

437 lines (300 loc) · 14.8 KB
Raw

Skills Catalog

All pre-built skills included in this template — what they do, how to invoke them, and which agents support them.

How Skills Work

Skills are invocable AI workflows. You call them by name in your agent's chat interface and the agent follows the skill's procedure — no extra setup required.

Not sure which skill to use?

flowchart TD
    Start(["What are you doing?"])

    Start --> Q1{"Exploring an idea\nor new feature?"}
    Q1 -->|Yes| Brainstorm["/brainstorming"]
    Q1 -->|No| Q2{"Hit a bug or\nunexpected behavior?"}
    Q2 -->|Yes| Debug["/systematic-debugging"]
    Q2 -->|No| Q3{"Got requirements,\nneed a plan?"}
    Q3 -->|Yes| Plan["/writing-plans"]
    Q3 -->|No| Q4{"Building\nUI / frontend?"}
    Q4 -->|Yes| FE["/frontend-design"]
    Q4 -->|No| Q5{"Task complete —\nready to review?"}
    Q5 -->|Code review| Review["/requesting-code-review"]
    Q5 -->|Feature done| Delivery["/feature-delivery"]
    Q5 -->|Urgent fix| Hotfix["/hotfix"]
    Q5 -->|Pre-release| Release["/release-readiness"]
    Q5 -->|Docs drifting?| DocAudit["/doc-audit"]
    Q5 -->|Security check| QA["/qa-audit"]
Loading
Agent How to invoke
Claude Code /skill-name in chat
GitHub Copilot @workspace /skill-name or reference in prompt
Cursor Mention skill name in chat
Kilo Code /skill-name in chat
Antigravity / Gemini Mention skill name; agent auto-discovers from .agent/skills/

Skills live in three parallel directories (same content, different paths for agent compatibility):

  • .claude/skills/ — Claude Code
  • .agents/skills/ — Universal (Copilot, Codex, Cursor, Kilo Code)
  • .agent/skills/ — Antigravity / Gemini

Source legend

Each skill entry is tagged with its origin:

Tag Source
[obra/superpowers] github.com/obra/superpowers — by Jesse Vincent (@obra)
[skills.sh] skills.sh marketplace — third-party verified skill
[this template] Created for this template — contains [FILL:] markers for your project

Installing skills from skills.sh

Browse skills at skills.sh, then install in three steps:

1. Find the skill URL on skills.sh (each skill has a raw SKILL.md link).

2. Install into all three directories:

SKILL=my-skill-name
SKILL_URL=https://skills.sh/skills/my-skill-name/SKILL.md

mkdir -p .claude/skills/$SKILL .agents/skills/$SKILL .agent/skills/$SKILL
curl -sL "$SKILL_URL" -o .claude/skills/$SKILL/SKILL.md
cp .claude/skills/$SKILL/SKILL.md .agents/skills/$SKILL/SKILL.md
cp .claude/skills/$SKILL/SKILL.md .agent/skills/$SKILL/SKILL.md

3. Verify:

ls .claude/skills/$SKILL/SKILL.md   # should exist

The skill is immediately available — no restart needed. Invoke it in your agent with /my-skill-name.

Installing the full obra/superpowers suite (recommended):

git clone https://github.com/obra/superpowers /tmp/superpowers
for dir in .claude/skills .agents/skills .agent/skills; do
  cp -r /tmp/superpowers/skills/* "$dir/"
done

Setup Skill

setup-repo [this template]

Run immediately after cloning — analyze the codebase and auto-generate filled versions of every agent instruction file.

Reads package.json, config files, git hooks, CI config, and source structure to infer project name, tech stack, constraints, prohibitions, test commands, and file structure. Outputs ready-to-paste content for all 8 instruction files, replacing every [FILL:] marker.

Works without any manual input — the agent derives everything from the codebase itself.

For non-Claude Code agents: run npx repomix first to bundle the repo, then attach the output.
Invoke /setup-repo
When Immediately after cloning; when manual fills aren't deep enough
When NOT Repo already fully set up (no [FILL:] markers remain)
Needs [FILL:] No
Agents All
Copilot /setup-repo prompt in .github/prompts/setup-repo.prompt.md

Universal Skills

These skills need no project customization. Use them as-is.

brainstorming [this template]

Use before any creative work — features, components, UI changes, new functionality.

Explores user intent, design constraints, and edge cases through structured dialogue before writing a single line of code. Prevents building the wrong thing.
Invoke /brainstorming
When Before implementing anything non-trivial
When NOT For trivial one-liner fixes or purely mechanical tasks
Needs [FILL:] No
Agents All

systematic-debugging [obra/superpowers]

Use before proposing any fix to a bug, test failure, or unexpected behavior.

Enforces root-cause tracing instead of guessing. Follows a structured investigation: reproduce → isolate → hypothesize → verify → fix. Resists pressure to jump to conclusions.
Invoke /systematic-debugging
When Any bug, failing test, or "why is this broken?" question
When NOT When you already know the root cause with certainty
Needs [FILL:] No
Agents All

writing-plans [this template]

Use when you have requirements for a multi-step task, before touching code.

Produces a detailed implementation plan: which files to touch, in what order, how to test each step, and what to verify. Generates bite-sized tasks. Assumes zero context from the engineer reading it.
Invoke /writing-plans
When Before starting any multi-step feature or refactor
When NOT Single-file changes that fit in one prompt
Needs [FILL:] No
Agents All

requesting-code-review [this template]

Use after completing a task or before merging, to catch issues early.

Dispatches a qa-reviewer subagent (Claude) that reviews your changes against the project's rules and surfaces violations before they become problems.
Invoke /requesting-code-review
When After implementing a feature, before creating a PR
When NOT During development — wait until a logical checkpoint
Needs [FILL:] No
Agents Claude Code (subagent dispatch), others (inline review)

skill-creator [skills.sh]

Use when you want to add a new skill to the library or update an existing one.

Guides you through the skill format, frontmatter conventions, workflow structure patterns, and cross-agent compatibility requirements.
Invoke /skill-creator
When Creating or updating a skill
When NOT When editing regular project files
Needs [FILL:] No
Agents All

find-skills [obra/superpowers]

Use when you want to discover skills from the open agent skills ecosystem.

Helps you find skills that might exist for your task — both in this repo and from the broader community. Covers installation and integration.
Invoke /find-skills
When "Is there a skill for X?", "How do I do Y with an agent?"
When NOT When you know exactly what you need
Needs [FILL:] No
Agents All

using-superpowers [obra/superpowers]

Meta-skill — establishes the rule that skills must be invoked before responding to any request where they apply. Activates at conversation start to prime the agent.
Invoke /using-superpowers (or load at start of session)
When Start of any session where you want maximum skill usage
Needs [FILL:] No
Agents All

frontend-design [skills.sh]

Use when building UI — web components, pages, dashboards, HTML/CSS layouts, React components.

Generates production-grade frontend code with strong aesthetic judgment. Deliberately avoids generic "AI slop" patterns. Prioritizes visual distinctiveness and working code over boilerplate.
Invoke /frontend-design
When Any UI/frontend task
When NOT Backend, config, or infrastructure tasks
Needs [FILL:] No
Agents All

jules-tasks [this template]

Use to audit and manage Jules AI sessions.

Lists active Jules sessions, reviews their status, and helps batch-manage tasks delegated to Jules.
Invoke /jules-tasks
When Reviewing Jules batch work, checking session status
Needs [FILL:] Requires JULES_API_KEY in .env
Agents Claude Code, Antigravity

Project-Specific Skills [this template]

These skills contain [FILL:] markers. Fill them in for your project before using.

qa-audit [this template]

9-phase security and robustness audit of your codebase.

Covers: XSS/injection scanning, rule compliance verification, race condition analysis, config schema validation, and documentation drift detection. Produces a severity-classified report (Critical / High / Medium / Low).
Invoke /qa-audit
When Before releases, after large refactors, when reviewing security posture
Needs [FILL:] Yes — project-specific scan patterns in SKILL.md
Agents All (allowed-tools restricted to read/grep for safety)

doc-audit [this template]

8-phase documentation drift audit across all docs, agent instructions, and config files.

Checks that every count, path, file reference, and config field in documentation matches the actual codebase. Surfaces stale docs, broken links, and inconsistencies.
Invoke /doc-audit
When After adding/removing files, before releases, when docs feel out of date
Needs [FILL:] Yes — file paths and expected counts for your project
Agents All (disable-model-invocation: true — deterministic scan only)

feature-delivery [this template]

End-to-end feature workflow: planning → implementation → QA → docs sync.

Follows .agent/workflows/feature-delivery.md strictly. Ensures every feature goes through the full delivery checklist: design, implementation, tests, documentation update, and review.
Invoke /feature-delivery
When Delivering a new feature or larger enhancement
When NOT Small bug fixes (use /hotfix instead)
Needs [FILL:] Yes — workflow file has project-specific steps
Agents All

hotfix [this template]

Fast but controlled workflow for urgent production fixes.

Prioritizes speed while maintaining safety: targeted scope, mandatory test verification, no scope creep. Follows .agent/workflows/hotfix.md.
Invoke /hotfix
When Critical bug in production that needs an immediate fix
When NOT Non-urgent issues — use /feature-delivery
Needs [FILL:] Yes — deployment/rollback steps in workflow file
Agents All

release-readiness [this template]

Pre-release quality gate: functional checks, baseline validation, documentation integrity.

Runs through a checklist of everything that must be true before tagging a release: tests pass, docs are synced, no outstanding [FILL:] markers, CHANGELOG is updated.
Invoke /release-readiness
When Before publishing a new version or tagging a release
Needs [FILL:] Yes — release checklist items for your project
Agents All

architecture-change [this template]

Controlled workflow for structural/architectural changes with mandatory ADR.

Requires writing an Architecture Decision Record (ADR) before implementing. Enforces controlled rollout, documentation update, and risk assessment.
Invoke /architecture-change
When Any change that affects the system's structure, data flow, or core dependencies
When NOT Feature additions that don't change architecture
Needs [FILL:] Yes — ADR template and review process in workflow file
Agents All

Prompt Files (GitHub Copilot)

These are Copilot-specific prompt files triggered with /command in VS Code Copilot Chat:

Prompt Invoke What it does
new-feature.prompt.md /new-feature Structured feature planning — explores requirements before implementation
review-changes.prompt.md /review-changes Pre-commit rule compliance check — catches violations before committing

Prompt files live in .github/prompts/.

Adding New Skills

Use the /skill-creator skill to guide creation. The format is:

.claude/skills/my-skill/
    SKILL.md          ← Required: frontmatter + instructions
    references/       ← Optional: supporting docs

Copy to all three skill directories to make it available across agents:

cp -r .claude/skills/my-skill .agents/skills/my-skill
cp -r .claude/skills/my-skill .agent/skills/my-skill

Skills Ecosystem — Discover More

The skills in this template are a starting point. The broader ecosystem has hundreds more.

Where to find skills

Resource What's there
skills.sh Public marketplace — browse, install, and publish skills
github.com/obra/superpowers The original superpowers suite by Jesse Vincent — using-superpowers, find-skills, systematic-debugging, and more
github.com/google-labs-code/jules-awesome-list Curated Jules prompts and workflows

Installing a skill from skills.sh

# Install via the skills CLI (if available)
skills install <skill-name>

# Or manually: copy SKILL.md into the correct directory
mkdir -p .claude/skills/<skill-name>
curl -o .claude/skills/<skill-name>/SKILL.md <skill-url>

# Then mirror to the other skill directories
cp -r .claude/skills/<skill-name> .agents/skills/<skill-name>
cp -r .claude/skills/<skill-name> .agent/skills/<skill-name>

Installing obra/superpowers

The complete superpowers suite (including skills not in this template):

# Clone and copy all superpowers skills
git clone https://github.com/obra/superpowers /tmp/superpowers
cp -r /tmp/superpowers/skills/* .claude/skills/
cp -r /tmp/superpowers/skills/* .agents/skills/
cp -r /tmp/superpowers/skills/* .agent/skills/

Attribution when adding external skills

When adding community skills, update this catalog with the source tag so it's clear what came from where. The [FILL:] marker convention only applies to skills created for this template — external skills should be used as-is and updated by pulling from their upstream repo.