Prod-Readiness

Spawns 5 specialist review agents in parallel, each writing findings to a dedicated report file. Synthesizes into a prioritized production readiness report with executive summary, blockers, and remediation plan.

Do NOT load when

• user wants only one narrow review (use the relevant specialist skill instead)
• codebase is mid-implementation with a knowingly broken baseline
• task is to fix issues rather than assess readiness

NEVER

• Never run on a dirty working tree — uncommitted changes mean agent findings don't map to a stable commit; findings become unreproducible and the report loses traceability
• Never run on a failing baseline — a red test suite makes it impossible to distinguish pre-existing failures from review findings; always confirm baseline is green first
• Never synthesize before all 5 agents finish writing — partial synthesis produces a report that omits entire dimensions; a CRITICAL finding from a slow agent gets buried in LOW backlog
• Never let agents write to the same file — each agent has its own REVIEW_*.md; shared files produce interleaved, unparseable output
• Never commit REVIEW_*.md files from a previous run without clearing them — stale findings from a prior session mixed with new ones produce a misleading severity distribution
• Never rate an issue CRITICAL without a file:line reference — untraceable CRITICALs create review fatigue and get deprioritized the same as vague LOWs