A collection of reusable skills for AI coding agents (Claude Code, Codex, and others).
Each skill is a self-contained directory with a SKILL.md that defines when and how the agent should use it.
| Skill | Description |
|---|---|
| codebase-audit | Deep codebase audit: understand the project first, then systematically find hardcoded constants and unfinished code |
| cli-for-agent | Design and review CLIs for reliable agent automation: non-interactive flags, structured output, fast errors, idempotency |
| consult-chatgpt-pro | Consult ChatGPT Pro through a desktop browser with sanitized review packets, action-time approval, and local validation |
| consult-chatgpt-pro-browser | Consult ChatGPT Pro through the Codex in-app browser with sanitized review packets, action-time approval, send verification, and local validation |
| init | Generate a minimal AGENTS.md context file for a repository using the WHAT/WHY/HOW framework |
| mcp-for-agents | Design and review MCP servers for AI agents: outcome-oriented tools, flat parameters, actionable errors, token efficiency |
| systematic-fix | Fix-verify-lockdown workflow: catalog issues, fix one at a time with verification, then write e2e regression tests covering each fix and its failure family |
| simplify-code | Review git diffs for reuse, quality, efficiency, and clarity issues with parallel sub-agent reviews, then optionally apply safe behavior-preserving fixes |
| blueprint | Generate visual HTML blueprint pages and structured markdown plan specs with architecture diagrams, phased task breakdowns, file change maps, dependency graphs, risk matrices, and verification checklists |
| visual-explainer | Generate self-contained HTML visualizations for architecture diagrams, flowcharts, data tables, timelines, and dashboards |
| bug-hunt-swarm | Parallel read-only multi-agent root-cause investigation that ranks hypotheses, verifies the leading one against real code, and recommends the fastest proof path |
| review-swarm | Parallel read-only multi-agent review of a diff or file scope, surfacing regressions, security risks, performance issues, and contract gaps with a prioritized fix path |
| orchestrator-only | Orchestration-only mode: brainstorm, spec, decompose, and dispatch sub-agents in parallel without implementing the work yourself |
| next-move | Deep strategic analysis to identify the single highest-leverage, most innovative addition for a project |
| goal-distill | Compress the current session into a paste-ready /goal prompt that drives an autonomous phase loop in a fresh session |
Install any skill from this collection using the skills CLI:
npx skills add ReinaMacCredy/Skills# Install a single skill
claude skill add /path/to/Skills/systematic-fix
# Or symlink into your skills directory
ln -s /path/to/Skills/systematic-fix ~/.claude/skills/systematic-fixCopy or symlink the skill directory into your agent's skill search path.
Each skill follows this layout:
skill-name/
SKILL.md # Required. Frontmatter (name, description) + instructions.
reference/ # Optional. Docs loaded into context as needed.
templates/ # Optional. File templates used in output.
prompts/ # Optional. Sub-agent prompt templates.
See individual skill directories for licensing. Skills without explicit licenses are provided as-is.