agentleFS
lowtidebuildlowtidebuild/second-review-agent/.claude/skills/scoring-engine/SKILL.md

scoring-engine

Skill1 starsChanged 7 months agoApache-2.0View on GitHub

Compute per-dimension review scores, overall grade, and independent release recommendation. Grade measures overall quality. Release recommendation is a safety gate. They are independent: - Grade A + "Manual Review Required" = excellent analysis but one key citation unverified → human must verify - Grade C + "Pass with Warnings" = many minor issues but no substance problems → may release with awareness The release recommendation is the binding safety decision. The grade is informational.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Skipped Dimension Handling: If a dimension is skipped (e.g., Dim 3 when no client context), set score: null, skipped: true, skip_reason: …

    line 15
  2. Distinctive:

    Pass with Warnings: No Dim 1–3 Criticals, but Majors exist in any dimension.

    line 21
  3. Distinctive:

    Pass: No Critical or Major findings.

    line 22
  4. Very specific:

    After issue consolidation, scoring, and known-issues matching complete, the main agent updates checkpoint.json:

    line 94

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

scoring-engine Skill

Compute per-dimension review scores, overall grade, and independent release recommendation.

Capabilities

  1. Per-Dimension Scoring (LLM judgment)
    • Score each dimension 1–10 based on findings
    • Scale: 10=no issues, 7–9=minor only, 4–6=major present, 1–3=critical found
    • Reference scoring-rubric.md for detailed criteria
  2. Overall Grade Computation
    • Average of all applicable dimension scores
    • A (avg ≥ 8.5), B (avg ≥ 7.0), C (avg ≥ 5.0), D (avg < 5.0)
    • Skipped Dimension Handling: If a dimension is skipped (e.g., Dim 3 when no client context), set score: null, skipped: true, skip_reason: "..." in scorecard. Exclude from average: overall_average = sum(non-null scores) / count(non-null scores). Skipped dimensions are NOT treated as 0.
    • Grade Constraint: Grade MUST be exactly one of A, B, C, D. No plus/minus modifiers (no "C+", "B-", etc.)
  3. Release Recommendation (independent safety gate)
    • Release Not Recommended: Any Dim 1–3 Critical finding; OR any Nonexistent citation on dispositive conclusion
    • Manual Review Required: Any Unverifiable citation on key conclusion; OR Dim 2 has ≥2 Major findings
    • Pass with Warnings: No Dim 1–3 Criticals, but Majors exist in any dimension; OR grade < B
    • Pass: No Critical or Major findings; grade ≥ B
    • Constraint: release_recommendation MUST be exactly one of these four string values. Do NOT use free-form text. Use release_rationale for the explanation.
    • Evaluation order: top-to-bottom, first match wins:
      if any Critical in Dim 1–3 OR any Nonexistent on dispositive conclusion → "Release Not Recommended"
      elif any Unverifiable on key conclusion OR Dim 2 Major count >= 2 → "Manual Review Required"
      elif no Dim 1–3 Criticals AND (any Major OR grade < B) → "Pass with Warnings"
      else → "Pass"
  4. Issue Consolidation
    • Merge all findings from Steps 3–5 into issue-registry.json
    • Deduplicate (same location + same description = merge)
    • Priority sort: (1) severity descending, (2) dimension order, (3) document order
    • Tag known-issue matches via known-issues-manager skill

When to Use

  • WF1 Step 6: Issue Consolidation & Scoring
  • Input: All findings from Steps 3–5, verification-audit.json, review-manifest.json
  • Output: issue-registry.json, review-scorecard.json
  • Script: scripts/assemble-review-output.py
  • Usage: python3 assemble-review-output.py <working_dir> [--legacy-issue-registry <path>] [--legacy-scorecard <path>]

Grade vs. Release Interaction

Grade measures overall quality. Release recommendation is a safety gate. They are independent:

  • Grade A + "Manual Review Required" = excellent analysis but one key citation unverified → human must verify
  • Grade C + "Pass with Warnings" = many minor issues but no substance problems → may release with awareness

The release recommendation is the binding safety decision. The grade is informational.

Output Schemas

issue-registry.json

{
  "issues": [
    {
      "issue_id": "ISS-001",
      "dimension": 1,
      "severity": "Critical",
      "location": {"section": "III.2", "paragraph_index": 45},
      "description": "...",
      "recommendation": "...",
      "evidence": {"citation_id": "CIT-003"},
      "recurring_pattern": null
    }
  ]
}

review-scorecard.json

{
  "dimensions": {
    "1_citation": {"score": 6, "skipped": false, "findings_count": {"Critical": 1, "Major": 0, "Minor": 2}},
    "2_substance": {"score": 8, "skipped": false, "findings_count": {}},
    "3_alignment": {"score": null, "skipped": true, "skip_reason": "No client context provided", "findings_count": {}},
    "4_writing": {"score": 7, "skipped": false, "findings_count": {}},
    "5_structure": {"score": 9, "skipped": false, "findings_count": {}},
    "6_formatting": {"score": 8, "skipped": false, "findings_count": {}}
  },
  "overall_grade": "B",
  "overall_average": 7.6,
  "release_recommendation": "Manual Review Required",
  "release_rationale": "Dim 1에 Critical finding 1건 (CIT-003: Nonexistent citation)"
}

Checkpoint

After issue consolidation, scoring, and known-issues matching complete, the main agent updates checkpoint.json:

  • step_6.status → "completed"
  • step_6.output → "working/issue-registry.json,working/review-scorecard.json"
  • last_completed_step → 6

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.