Skip to content

Skill Catalog

Every entry below is stored once at skills/<category>/<name>/SKILL.md and can be discovered by Claude Code, Codex, or Hermes through their normal skill-loading mechanisms. The category headings match the directory that groups each skill on disk.

orchestration

  • orchestrator-workflow — coordinate complex implementation work through discovery, planning, execution, and validation.
  • refactor-workflow — perform controlled refactoring with explicit evidence and verification.
  • agent-failure-recovery — stop retry loops, re-observe world state, run one discriminating probe, and recover containably.
  • agent-harness-design — design safe, diagnosable tools, schemas, observations, permissions, and recovery contracts for autonomous agents.

skill-management

repository

software-development

  • architecture-decision-records — record and maintain significant architecture decisions with grounded alternatives, consequences, and revisit triggers.
  • interaction-state-audit — trace UI interactions through shared state, effects, and asynchronous writers to explain incorrect final state.
  • ai-assisted-regression-testing — counter shared AI author-review assumptions with executable regression oracles across important boundaries.
  • browser-qa — collect safe browser-based release evidence for runtime health, interactions, visual baselines, and accessibility.

devops

games

Provider-specific references are included only when they document public contracts or reusable failure modes. Personal, account, infrastructure, and private-project evidence is excluded.