A collection of Claude Code custom skills and agent definitions that enforce disciplined, worker-routed AI-assisted development workflows.
概要
A collection of Claude Code custom skills and agent definitions that enforce disciplined, worker-routed AI-assisted development workflows. If you like what I do, consider supporting my work via donation Reusable process guides that Claude Code invokes via slash commands. Each skill lives in its own directory with a SKILL.md file. Worker agent definitions that skills route tasks to. Each agent has a focused role and strict boundaries. Copy skills and agents into your Claude Code configuration directory: Then add the routing instruction to your ~/.claude/CLAUDE.md: 1. Every task triggers using-superpowers first 2. The skill selects an ordered pipeline of worker agents (e.g. task-planner -> feature-implementer -> test-engineer -> code-reviewer) 3. Each worker self-identifies, stays in role, and produces structured output 4. No silent fallback to generic behavior — routing is always visible
README
Skills & Agents
A collection of Claude Code custom skills and agent definitions that enforce disciplined, worker-routed AI-assisted development workflows.
If you like what I do, consider supporting my work via donation
What’s Inside
Skills (skills/)
Reusable process guides that Claude Code invokes via slash commands. Each skill lives in its own directory with a SKILL.md file.
| Skill | Purpose |
|---|---|
using-superpowers |
Router/gatekeeper — routes every task through the correct worker agent pipeline before any work begins |
writing-plans |
Structured implementation planning before coding |
test-driven-development |
RED-GREEN-REFACTOR cycle enforcement |
systematic-debugging |
Root-cause analysis before proposing fixes |
brainstorming |
Explores intent, requirements, and design before implementation |
interview |
In-depth user interviewing to create detailed specs |
writing-skills |
TDD-based approach to creating and testing new skills |
find-bugs |
Security and code quality review of local branch changes |
code-simplifier |
Refines code for clarity and consistency while preserving behavior |
requesting-code-review |
Verify work meets requirements before merging |
receiving-code-review |
Technical rigor when processing review feedback |
verification-before-completion |
Evidence-based verification before claiming work is done |
dispatching-parallel-agents |
Parallelizes independent tasks across subagents |
subagent-driven-development |
Executes plans with independent implementation agents |
executing-plans |
Runs implementation plans with review checkpoints |
finishing-a-development-branch |
Guides branch completion — merge, PR, or cleanup |
Agents (agents/)
Worker agent definitions that skills route tasks to. Each agent has a focused role and strict boundaries.
| Agent | Role |
|---|---|
task-planner |
Translates tasks into minimal, executable implementation plans |
feature-implementer |
Implements one scoped plan step with minimal diff |
test-engineer |
Writes/updates tests, regression coverage, edge cases |
code-reviewer |
Reviews correctness, regressions, standards, test adequacy |
acceptance-checker |
Maps implementation to acceptance criteria with evidence |
docs-updater |
Keeps documentation aligned with recent code changes |
api-researcher |
Researches external APIs/libraries and produces guidance |
bug-repro-triager |
Produces repro steps, hypotheses, and investigation plans |
code-simplifier |
Simplifies recently modified code while preserving behavior |
research |
General-purpose research agent |
Installation
Copy skills and agents into your Claude Code configuration directory:
# Skills
cp -r skills/* ~/.claude/skills/
# Agents (if supported by your setup)
cp -r agents/* ~/.claude/agents/
Then add the routing instruction to your ~/.claude/CLAUDE.md:
For EVERY task — no exceptions — invoke the `using-superpowers` skill before doing any work.
How It Works
- Every task triggers
using-superpowersfirst - The skill selects an ordered pipeline of worker agents (e.g.
task-planner -> feature-implementer -> test-engineer -> code-reviewer) - Each worker self-identifies, stays in role, and produces structured output
- No silent fallback to generic behavior — routing is always visible
License
MIT
推奨ツール
別のキーワードを試すか、フィルタを外してください。
インストール
npx skillfish add roman01la/skills-agents