archive-sessionlisted

You are archiving the session that is ending. Follow the **session-archive-conventions** memory (folder taxonomy, file naming, metadata-block format) exactly. The mechanical half is a script; you do the interpretive half from what you already lived this session — you should NOT need to re-read the whole transcript. ## Step 1 — Identify the target project - If `$ARGUMENTS` names a project or path, use it. Otherwise infer the project this session actually worked on (the repo you edited files in — which may differ from the launch cwd). If genuinely ambiguous, ask the user once; otherwise proceed. - Resolve its root (the dir containing `private/`). Confirm `private/` with the taxonomy folders exists; if not, create `private/{transcript,findings,decisions,session_summaries,origin}/`. ## Step 2 — Classify the session Decide, honestly: - **Topic slug** — kebab-case, ≤5 words, describing what the session covered. - **High-signal?** Per the convention, a metadata block + findings + decisions are ONLY for high-signal sessions (origin, major pivot, first benchmark, breakthrough, major failure). Routine sessions (refactors, pushes, small fixes) get a transcript + summary only — no metadata block, no findings/decisions. ## Step 3 — Run the mechanical script Run (substitute real values; `--date` is today's date from your context): ``` python3 ~/.claude/scripts/session_to_markdown.py \ --project-root "<ROOT>" --topic-slug "<SLUG>" --date "<YYYY-MM-DD>" ``` It writes the transcript body