imtemp-dev

bts-adjudicate

Evaluate a debate conclusion for feasibility, over-engineering, and evidence quality. Decides whether to accept the conclusion or drive another debate round with targeted preparation (web search, code analysis, repo exploration).

bts-architect

Propose at least 2 alternative structural decompositions, debate them, adjudicate, and commit the winner to wireframe.md. Prevents the single-path trap where the first decomposition the model imagines becomes the only one considered.

bts-assess

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.

bts-audit

Audit a document for completeness. Find missing scenarios, unconsidered edge cases, and hidden assumptions. Includes mermaid branch completeness analysis. Use after verify and cross-check.

bts-comment-apply

Incorporate inline BTS review comments embedded in a recipe's docs. Reads `.bts/local/recipes/{id}/pending-comments.json` (written by `bts comment apply`), runs meta-analysis (conflicts, cascades, verify-contradictions, ambiguity), applies edits per-doc in parallel, and shows a git diff for approval before finalizing.

bts-debate

Run a structured expert debate with 3 personas. Produces a decision document with rationale. State is saved for resuming later.

bts-discover

Explore the intent and direction behind a feature request through multi-turn conversation. Clarifies purpose, users, problem, and direction before scoping begins.

AI & Automation Listed

bts-domain-model

Model the problem domain BEFORE structural decomposition. Defines entities, invariants (each with a single owner), state partitioning (transient/persistent/derived), and enumerates legal/illegal combinatorial state cells. Produces domain.md as the contract /bts-wireframe and /bts-architect must honor.

bts-implement

Implement code from a finalized Level 3 spec (final.md). Uses an adaptive loop with build verification — the same ASSESS→action→VERIFY pattern as spec creation.

bts-recipe-analyze

Analyze an existing system or codebase. Produces a verified Level 1 (understanding) document.

bts-recipe-blueprint

Create a Level 3 implementation spec through an adaptive loop of research, drafting, debate, simulation, and verification. The loop continues until the document is bulletproof.

bts-recipe-debug

Debug unknown bugs through multi-perspective analysis. Collects 6 "blueprints" (data flow, dependencies, design intent, runtime, change history, impact), cross-references them to find root cause, runs adversarial validation on the top hypothesis, then produces a verified fix spec implementable via /bts-implement.

bts-recipe-design

Design a feature or system. Produces a verified Level 2 (design) document. Can be followed by /recipe blueprint to reach Level 3.

bts-recipe-fix

Diagnose and fix a bug through document-first approach. Creates fix-spec.md with root cause analysis, adversarial validation, simulation, expert review, and verified implementation. Lighter than blueprint — no scoping, no task decomposition, 1-round debate.

bts-research

Systematically research code, documentation, or external sources. Produces a structured research document. Use at the start of any recipe.

bts-review

Multi-perspective code review with quality, security, and architecture agents. Basic review (all perspectives) by default, focused review with category, or lightweight mid-run checkpoint during implementation.

bts-simulate

Walk through scenarios to find gaps and incorrect assumptions. Document mode: test a spec document. Code mode: test implemented code against its spec. Both use scenario-based walkthrough.

bts-status

Generate or update project-status.md — a comprehensive view of all recipes, their implementation state, deviations, and next steps.

bts-sync

Synchronize final.md spec with actual implementation. Finds deviations between spec and code, updates final.md to reflect reality, and generates deviation report.

Testing & QA Listed

bts-test

Generate and run tests based on final.md test scenarios. Runs an adaptive loop: execute tests, analyze failures, fix implementation code, re-run until all pass.

bts-verify

Verify a document for logical errors, contradictions, and unsupported claims. Includes mermaid flow path enumeration to find unspecified execution paths.

Data & Documents Listed

bts-wireframe

Design the high-level structure of a feature using mermaid diagrams. Creates wireframe.md with component diagram, state machine, data flow, file structure, and all execution paths enumerated.

bts-cross-check

Check internal consistency of a document — terms used consistently, interfaces match between sections, no contradictions. Optionally checks against source code if it exists.