Review implementation against architectural intent and produce findings. Use when the user wants to check whether what was built matches the architecture, identify drift, or assess architectural risks.
Review implementation against architectural intent. Produce findings.
The user may specify what to review: $ARGUMENTS
Read these files automatically:
docs/architecture.md — the architectural intent to evaluate againstdocs/project-brief.md — project constraints and decisions.workflow/handoff-notes/swe/ — what was built and what changedIf docs/architecture.md doesn't exist, tell the user: "No architecture document exists yet. Ask for the sa-design skill first to establish the architectural intent, or this review will have no baseline to evaluate against."
If the user specified what to review (e.g., a specific component, a recent session's work), focus there. Otherwise, review broadly.
State what you're reviewing:
For each area in scope, check:
Categorize findings:
For each finding:
For must-fix and should-fix findings, create issue files in .workflow/issues/backlog/. Run .cursor/scripts/next-issue-number.sh for the next available number.
[expert]-[type]-[number].mdRun .cursor/scripts/update-issues-list.sh after creating issues.
Present findings to the user. Highlight the most important items first.
If the architecture document itself needs updating (e.g., intentional drift should be documented), suggest asking for the sa-update skill.
Do NOT auto-fix implementation. Fixes go through the SWE workflow so they get proper testing and verification. Your job is to find problems, not fix them.