Assess a document's current level (1-3) and determine the next action needed. This is the brain of the adaptive loop — it decides what to do next.
Assess the document and decide the next action.
Run level assessment via bts binary:
bts verify $ARGUMENTS
This returns the current level score and missing criteria.
Build situational awareness:
Read the document fully yourself and evaluate:
Decide the next action based on assessment:
If information is insufficient → recommend /research
"Need to investigate [specific topic] before proceeding."
If technical decision is uncertain → recommend /debate
"The choice between [A] and [B] needs expert discussion."
If gaps may exist → recommend /simulate
"Walk through [specific scenarios] to find blind spots."
If content needs to be added → recommend IMPROVE "Add [specific items] to reach Level [N]. Then run /verify."
If contradictions are suspected → recommend /verify
"Check sections [X] and [Y] for consistency."
If completeness is uncertain → recommend /audit
"Review for missing error cases, edge cases, security."
If mermaid diagrams are missing or incomplete → recommend IMPROVE "Add state machine and flow diagrams with all execution paths enumerated."
If Level 3 criteria all met → recommend /sync-check then finalize
"Document appears complete. Run sync-check before finalizing."
Output your assessment:
## Assessment
Current Level: [X.Y]
Missing for next level: [list]
## Recommended Action
[ACTION]: [specific instruction]
## Rationale
[Why this action is needed now]
Priority rule: If verify-log shows critical=0 has been achieved AND no simulation has run yet (check changelog.jsonl for "simulate" action), strongly recommend /simulate BEFORE further IMPROVE cycles.
Rationale: structural verification (critical=0) means the spec is internally consistent. Now is the best time to test scenarios — simulation catches behavioral gaps (failure modes, race conditions, edge cases) that verification cannot find. Running simulation early prevents discovering these gaps late and needing additional rework iterations.