Now liveThe Skillselion MCP - thousands of ranked skills, loaded into your agent mid-task. No install.Get it →
hayeonhee avatar

Validated Plan Qa

  • Updated April 9, 2026
  • hayeonhee/validated-plan-qa

validated-plan-qa is a Claude Code plugin skill that runs a three-layer post-audit on validated-plan execution results so developers can detect self-graded passes and structural gaps before trusting deliverables.

About

validated-plan-qa is a Claude Code plugin that post-audits validated-plan execution with a three-layer QA engine. Layer 1 checks artifact structure and criteria drift; Layer 2 gives an independent agent only the user-intent summary and final output; Layer 3 compares chain and independent verdicts for dangerous mismatches. Invoke it after validated-plan reports success when you need an adversarial second opinion before shipping or merging.

  • Three layers: structure audit, blind independent verdict, divergence analysis
  • Scripts and rubrics (R-1~R-5) enforce consistent PASS, FAIL, and WARN criteria
  • Blocks chain framing bias by withholding intermediate plan and criteria from Layer 2
  • Archives reports under .omc/qa-archive with self-reinforcing rubric case logging
  • Event-driven missed-patterns loop promotes recurring gaps into checklist rules

Validated Plan Qa by the numbers

  • Data as of Jul 7, 2026 (Skillselion catalog sync)
/plugin marketplace add hayeonhee/validated-plan-qa

Add your badge

Show developers this marketplace is listed on Skillselion. Paste this into your README.

Listed on Skillselion
Last updatedApril 9, 2026
Repositoryhayeonhee/validated-plan-qa

How do you verify a validated-plan run truly succeeded when the same context that planned, executed, and scored also declared success?

Runs a 3-layer post-audit on validated-plan outputs—structure checks, blind independent verdict, and divergence analysis—to catch self-graded pass results.

Who is it for?

Developers who finished a validated-plan workflow and want independent structural checks plus a blind verdict before merge or release.

Skip if: Projects that never use validated-plan, or teams seeking live code review of diffs instead of plan-execution artifact audit.

What you get

A final QA report, independent verdict, layer-1 findings, and metadata archived under .omc/qa-archive with divergence flags when chain and independent judgments disagree.

Recommended Marketplaces

FAQ

Why is Layer 2 called independent?

A separate agent sees only clarify-result.md and final artifacts, not the plan, criteria, or execution reviews from the original chain.

What is the most dangerous divergence pattern?

Chain PASS with independent FAIL, which suggests the original workflow missed issues the blind review caught.

Does it fix validated-plan outputs automatically?

No. It audits, reports, archives findings, and can feed rubric and checklist improvements through the missed-patterns loop.

Testingtesting

This week in AI coding

Five minutes, every Monday - the tools, releases and tactics for developers.

unsubscribe anytime.