maicongambini/omp-harness-guide/harness-evaluator-rubric
Claude Code and agent skill by maicongambini/omp-harness-guide
- community
Checking that the SKILL.md file is still in the repository.
Builds evidence-backed rubrics before the Judge phase. Use when establishing criteria to assess task completion. Do NOT use for self-evaluating changes (use harness-quality-gate).
Install
npx vskill@latest install maicongambini/omp-harness-guide/harness-evaluator-rubricSKILL.md
The same SKILL.md was published earlier as maicongambini/opencode-harness-guide/harness-evaluator-rubric.
Harness Evaluator Rubric
Run after Execute + Validate and before claiming any verdict.
Inputs
Find the best available contract:
docs/harness/sprint-contract.md- Feature spec/task AC under
.specs/features/* - User-approved PREVC plan in current session
Also load available progress state, feature state, and startup path, plus the newest gate report (docs/harness/quality/*.json) and the router tier.
Two kinds of row
Measured rows are copied from `agent-os/quality-thresholds.json` verbatim — metric name, direction, threshold — and scored from the gate report, not by reading code. Never restate a number in prose here; a rubric with its own copy of a threshold becomes a second, drifting source of truth within a sprint.
Judgement rows are the task-specific part, built from the contract. That is where this skill adds value.
Rubric
| Criterion | Kind | Weight | Pass Condition | Score (1-5) | Evidence | Notes | |
|---|---|---|---|---|---|---|---|
| Functional completeness | judgement | high | All approved AC met | ||||
| Verification | measured | high | Gate exited 0 at the required mode | ||||
| Test strength | measured | high | mutation_kill_ratio meets threshold/baseline | ||||
| Regression | measured | high | regression_suite clean; bug fix has a new test | ||||
| Coverage floor | measured | medium | line_coverage / branch_coverage meet threshold | ||||
| Maintainability | measured | medium | cyclomatic_max / module_lines_max within threshold | ||||
| Boundaries | measured | medium | boundary_violations at threshold | ||||
| Scope discipline | judgement | high | Only approved scope changed | ||||
| Blind spots | judgement | high | docs/review.md items answered | ||||
| Handoff readiness | judgement | medium | State and handoff updated for fresh session |
Add task-specific judgement rows from the contract. Do not invent measured rows — a metric that is not in the thresholds file is not measured.
Scoring
- 5: fully met with evidence.
- 3: partially met with documented gap.
- 1: not met or no evidence.
- Average ≥4.0 and no 1-scores: Accept candidate.
- Any 1-score: Block candidate.
- A measured row whose tool is absent scores `unavailable`, not 5 and not 1. It does not enter
the average; it is listed as a gap. Scoring an unmeasured row as a pass is how a gate that measures almost nothing reads as green.
- Any red measured row caps the candidate at Revise, whatever the average says.
Do not self-approve. Send rubric and evidence to Judge.
Security
- Security scanPassed, 95 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingomp/agent/skills/harness-evaluator-rubric/SKILL.md
Popularity
- 7-day trend score
- 5
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/maicongambini/omp-harness-guide/harness-evaluator-rubric)