| name | spec-review |
| description | Architectural and technical review of a design spec. Spawns a subagent reviewer that verifies spec claims against live cluster state and deployed manifests, checks library/framework docs via Context7, and produces severity-ranked findings. Use when iterating on spec quality before implementation planning.
|
| argument-hint | <spec-file-path> |
| arguments | ["spec"] |
Spec Review
You are reviewing the spec at $spec.
Instructions
- Read the review brief at review-brief.md
- Read the full spec at
$spec
- Spawn a general-purpose subagent to perform the review. Brief it with:
- The full review brief contents
- The spec file path to review
- Instruction to use Context7 for library/framework validation
- Instruction to search
cluster/apps/ for deployed manifests to verify spec claims
- Instruction to use
kubectl CLI to check live cluster state where manifests alone are insufficient
- Instruction to cross-reference findings against the spec's own risks table, phase checklists, and tradeoff callouts before flagging — do not raise concerns the spec already addresses
- After receiving findings, present results organized by severity
- Offer to update the spec based on findings
Rules
- NEVER append review notes, summaries, or review round logs to the spec file. The spec is the design document — not a review journal. Findings go in conversation output only.
- Only edit the spec to fix actual issues identified in findings (incorrect claims, missing sections, design gaps). Each edit should improve the spec's content, not add meta-commentary about the review process.