bb-review（BB 评论）

User Input

$ARGUMENTS

You MUST consider the user input before proceeding (if not empty).

Goal

Run a BuildBetter-enriched review for the current feature with one of these modes:

• usability: user-story walkthrough and UX validation
• code: implementation/code quality review
• both (default): run both

This command is BuildBetter-first: prioritize buildbetter-context.md, buildbetter-context.json, and user-stories.md as review evidence when available.

Execution Flow

1. Initialize context

- Locate the current feature directory and verify available documents. - Required docs: spec.md, plan.md, tasks.md. - BuildBetter docs: buildbetter-context.md, buildbetter-context.json, user-stories.md (optional but preferred). - If required docs are missing: stop and instruct the user to run the missing prerequisite skill (e.g., use the bb-specify, bb-plan, or bb-tasks skill first).

1. Resolve review mode

- Parse $ARGUMENTS: - if contains usability -> usability-only - if contains code -> code-only - if contains both or empty -> both - If arguments are ambiguous, default to both.

1. Load minimal review context

- From spec.md: user stories, acceptance scenarios, functional and non-functional requirements. - From plan.md: architecture constraints, intended UX/platform surfaces, technical tradeoffs. - From tasks.md: story coverage, dependency ordering, test tasks. - From BuildBetter artifacts (if present): exact quotes, evidence IDs, themes, customers affected, evidence-backed stories.

1. Usability review (if selected)

- Evaluate each primary user story through this narrative: - Is this easy to use? - Is this intuitive? - Does this solve the problem? - Does this work? - For each question, produce a verdict: pass, risk, or fail with rationale. - Use BuildBetter evidence IDs (BB-EXTRACTION-*) where available. - Detect whether this is a visual/UI-heavy feature (pages, flows, forms, dashboard, mobile/web UI states). - If visual/UI-heavy: - Recommend Playwright validation and provide a focused scenario list tied to user stories. - Include coverage for happy path, alternate flow, and failure/edge case. - If not visual-heavy: - Recommend API/integration behavior checks aligned with user stories and acceptance criteria.

1. Code review (if selected)

- Review implementation state using current branch changes and task intent. - Prioritize findings in this order: - correctness/behavioral regressions - security/privacy issues - reliability/performance risks - missing tests for user-story acceptance criteria - maintainability issues that materially increase delivery risk - Keep findings concrete with file references and impact statements. - If no issues are found, state that explicitly and list remaining testing gaps.

1. Coverage and traceability checks

- Map each reviewed user story to: - acceptance criteria coverage - task coverage (tasks.md) - evidence linkage (BuildBetter IDs) when available - Flag any story with weak evidence, missing validation, or missing implementation coverage.

1. Produce review report # Review Report ## Review Scope - Mode: [usability|code|both] - BuildBetter Evidence: [available|partial|missing] ## Usability Review - Story:... - Easy to use: [pass|risk|fail] - rationale - Intuitive: [pass|risk|fail] - rationale - Solves the problem: [pass|risk|fail] - rationale - Works: [pass|risk|fail] - rationale - Evidence refs: [BB-EXTRACTION-###] ## Playwright Recommendation - Needed: [yes|no] - Why: - Suggested scenarios: ## Code Review Findings - [Severity] [File] Finding summary and recommendation ## Coverage Matrix - Story -> acceptance criteria -> tasks -> evidence refs ## Verdict - Ready / Ready with risks / Not ready - Blocking issues: - Suggested next step:

- Write FEATURE_DIR/review.md with this structure:

1. Report completion

- Return the path to review.md. - Summarize: - mode used - whether Playwright is recommended - top risks/blockers - readiness verdict

Rules

• Be explicit and deterministic; avoid vague review statements.
• Never fabricate evidence IDs or implementation status.
• Keep BuildBetter quotes redacted if re-quoted.
• Do not modify spec.md, plan.md, or tasks.md in this command.