bts-assesslisted

# Document Assessment Assess the document and decide the next action. ## Division of labor This skill runs in a FORK: the full document read in step 3 stays out of the main loop's context (repeated assess rounds were its largest context accumulator). Everything the loop needs comes back as your final output — the `<bts-decision>` block (Part A) plus the human-readable rationale (Part B). The ORCHESTRATOR parses the block and executes the action; you only decide. Bash in this fork is ONLY for read-only commands (`bts verify`, `bts recipe status`). Never run state-mutating bts commands (log, create, finalize, …) or write files from this fork. ## Steps 1. Run level assessment via bts binary: ```bash bts verify $ARGUMENTS ``` This returns the current level score and missing criteria. If `$ARGUMENTS` is empty, resolve the active recipe via `bts recipe status` and target its draft: `bts verify .bts/specs/recipes/{id}/draft.md` (`bts verify` requires exactly one file argument). 2. Build situational awareness: - Read changelog.jsonl (last 5 entries) to know what was just done - Check if simulation has run (look for "simulate" action in changelog) - Check if debates exist and whether conclusions are reflected in the draft - Read scope.md to keep boundaries in mind 3. Read the document fully yourself and evaluate: - What level is this document at? (1=understanding, 2=design, 3=implementation-ready) - What specific content is missing to