Curated Agent Guidance
A curated, rated catalog of skills, rules, workflows, hooks, subagents, and MCP servers for coding agents (Claude Code, Codex, and portable formats).
Read directly by agents via llms.txt or the agentic-catalog plugin skill.
Shortcuts: ⌘ / Ctrl K commands · / filter catalog · ↑↓ navigate · Esc reset
Rubric & Evidence Levels
- Human-focused score: seven grades from 0 to 4, weighted by practical value (25%), dependability (20%), cognitive ease (15%), human control (15%), adoption cost (10%), compatibility (10%) and upkeep (5%). Weighted total / 40 gives a score out of 10. Higher is better on every axis.
- Compare fairly: all active entries use the human-focused rubric. Score sorting puts complete reviews first, incomplete reviews next, and reviews with blockers last. Incomplete reviews have no total; blockers cannot be cleared by a high average. Check project fit and evidence separately.
- Evidence tiers: reviewed (curator read) → used (observed in session) → tested (trigger test) → evaluated (outcome eval).
- Source: original/adapted (authored here), vendored (pinned upstream copy), or linked (external reference).
