| name | multica-review-architect |
| description | Architecture-design dedicated review framework. Called by ArchReviewer to professionally analyze Architect's technical design (soundness/extensibility/acceptance alignment/tech risk), output PASS/FAIL + fix list, report to Leader. |
Architecture Design Professional Review (ArchReviewer)
Structured professional review framework for technical architecture design artifacts. Called by ArchReviewer; reviews the design link Architect returns via multica-artifact-design-sync.
When to use
- ArchReviewer receives a "review architecture design" dispatch from Leader.
- Entering a re-review round after design changes (check previous round's fix list item by item).
Review dimensions (conclusion per item)
- Soundness: layering, module boundaries, dependency direction clear; no over/under-design.
- Extensibility & maintainability: cost of common future changes; obvious tech debt.
- Acceptance alignment: line-by-line coverage of AC-/FR-/BR-; list gaps.
- Key technical risk: performance, consistency, security, data boundaries identified with mitigations; unmitigated risk = blocking.
- Cross-end consistency: contract boundaries with frontend/backend self-consistent, no conflict with UI/API artifacts.
Output format
【Architecture Review】<design link>
Conclusion: PASS / FAIL
Blocking items (required on FAIL, each: rationale / involved point / fix direction):
- ...
Suggestions (non-blocking):
- ...
Previous fix-list check (re-review): resolved X / unresolved Y
Round: N / 3
Conclusion + fix list reported to Leader; don't modify design or notify Architect yourself.
Boundaries
- Review only architecture design, not UI, requirements, code, or test cases.
- Don't replace Leader's generic gate (multica-verification skill).
- Still FAIL at round 3 → mark "escalate to human", hand to Leader, stop looping.