← ClaudeAtlas

usage-retrolisted

Builds a sampled 7 or 30 day usage retrospective: thread counts, correction rate, delegation ratio, trends, top actions. Use for run usage retro, weekly retro, monthly retro.
coltonbearden/brainrot · ★ 0 · AI & Automation · score 61
Install: claude install-skill coltonbearden/brainrot
# usage-retro ## Purpose Build a 7-day or 30-day retrospective of the user's Claude usage from their chat history: thread count, correction rate, delegation ratio, positive rate, and top topics — with trend deltas against a prior retro file and a top-3 actions table — written to a dated file the user can attach next period as the baseline. Every rate this skill produces is sampled via keyword probes over search, never exhaustively counted. Honest denominators are the identity of this skill: report what was measured, mark what was capped, and never estimate what the tools cannot see. ## Triggers Run this skill when the user says: - `run usage retro` - `usage retro` - `weekly retro` (forces the 7-day window) - `monthly retro` (30-day window) ## Preconditions - The past-chats tools `recent_chats` and `conversation_search` must be available. If either is missing, say the retro cannot run, and stop. - Scope follows where the skill runs: chats inside Projects are invisible from outside a Project. Run in a chat outside any Project for the general retro; running inside a Project yields a per-Project retro — label the output header accordingly. - There is no cross-session state. Trend diffing works only if the user attached a prior `RETRO-*.md` to this conversation; never reconstruct a baseline from memory. - The report must be written to `/mnt/user-data/outputs/` and presented, so the user can save it and attach it next period. ## Metric definitions | Metric | How measured