Skill Claude CodeCodex
Use when the user asks for an insights report over their coding sessions ("how am I using my agent", "/insights-style report", "what's working / where do sessions go wrong"). Runs the 3-layer insights pipeline over AgentGem's goldmine tools and ends at publishing high-outcome sessions as Gems.
Skill Claude CodeCodex needs its repo
Use when building or editing an AgentGem miniapp / mini-game — the sealed, single-file HTML app under /.agentgem/miniapps/. Covers the seal, the host capability protocol, consent, privacy, and the sizing traps. Also read before changing the miniapp platform itself (packages/play, the host, the Play console surfaces) …
Skill Claude CodeCodex
Use when rendering an AgentGem analysis report — a self-contained HTML readout built from a rubric evaluation's FACTS JSON (factors, findings, hygiene scores, per-session rows). Covers the facts-only rule, the embedded-JSON anti-hallucination seam, theming, and the editorial composition contract.
Skill Claude CodeCodex
Use when the user asks what keeps going wrong in their coding sessions, wants a quick retro on their agent habits, or asks how to make sessions smoother. Verifies detector findings against real transcripts before advising. For a full narrative report, use agentgem-insights instead.
Skill Claude CodeCodex
Use when the user wants to share/publish a Gem from their real usage. Drives scan → review → privacy gate → sign & publish via the agentgem-distill MCP tools.
Skill Claude CodeCodex
Use when the user wants to score their agent steering — "how well am I steering my agents", "what's my steering level", "run gemit". Runs the AgentGem gemit self-assessment — one npx command, a local HTML report, optional share.
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: