← ClaudeAtlas

afk-debrieflisted

Goes through a finished afk run — reads .afk/summary.md and the traces, sorts each non-green by cause (wrong gate, ticket too big, real failure), proposes what to fix and what to relaunch, and records afk's own defects in its docs/defects.md. Run on waking up, before merging. Triggers: /afk-debrief, "go through the run", "what failed last night?", "why is #48 red?".
Warshoow/afk.sh · ★ 3 · AI & Automation · score 69
Install: claude install-skill Warshoow/afk.sh
# /afk-debrief The run is over. This skill reads what it left behind, says **why** each ticket is not green, and proposes what to fix before relaunching. A red is not a verdict on the ticket: the gate may have been wrong, the worktree badly seeded, the model unavailable. Sorting that out is the whole job. ## 1 — The summary ```bash cat .afk/summary.md ``` No file → no run made it to the end; go to `.afk/<n>.out`. Columns: result, PR, attempt, model, context, cost, duration. Below the table: integration verdict, gates used, green rate on the 1st attempt, and the resume commands for the reds. ## 2 — Read the run before the tickets Four things are read on the whole, and a single one of them can explain every red at once: | What you see in the summary | What it says | |---|---| | **integration red while the tickets are green** | it is not N problems, it is one: the combination. `.afk/integration-verify.txt`, worktree kept in `.afk/wt/_integration` | | **"duplicate numbers"** | two branches took the same ADR or migration number. No gate can see it, git neither: renumber before merging. The number stays with whoever cites it most (`git grep -c` decides) | | **"same path created by several branches"** | two branches created the same file, with two APIs both of them right. Each compiles without the other: only the combination says so | | **"tickets from this run cited in the merged docs"** | a sentence may be in the future tense about something delivered ten minutes ago. No