26 skills, 7 agents, and PM frameworks for Claude Code. Your sparring partner, not your assistant.
Skill Claude CodeCodex
Build an LLM-as-Judge evaluator for one specific failure mode. Binary pass/fail only. Use when a failure mode requires interpretation (tone, faithfulness, relevance, completeness) and cannot be checked with code. Do NOT use when the failure can be checked with regex, schema validation, or execution tests. Do NOT use…
Skill Claude CodeCodex
Evaluate RAG pipeline retrieval and generation quality separately. Measure Recall@k, Precision@k, MRR, NDCG@k for retrieval. Assess faithfulness and relevance for generation. Use when the AI feature uses retrieval (search, knowledge base, document QA). Do NOT use for non-RAG AI features.
Skill Claude CodeCodex
Create diverse synthetic test inputs using dimension-based tuple generation. Use when bootstrapping an eval dataset, when real user data is sparse, or when stress-testing specific failure hypotheses. Do NOT use when you already have 100+ representative real traces (use stratified sampling instead).
Skill Claude CodeCodex
Systematic error analysis on real AI traces. Read traces, judge pass/fail, let failure categories emerge from data, compute failure rates, decide what to fix. Use when you have 50+ test cases or are seeing production failures. Do NOT use when you have fewer than 20 test cases (use /start-evals first).
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: