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
- yuzhaopeng-up/teleagent-skillstheir own version, first seen Jun 29, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- yuzhaopeng-up/teleagent-skills★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository15 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- citation-auditorSkill
- citation-checkerSkill
- cover-memo-writerSkill
- cross-document-checkerSkill
- document-parserSkill
- formatting-reviewerSkill
- ingestSkill
- known-issues-managerSkill
- library-managerSkill
- quality-gateSkill
- redline-generatorSkill
- structure-checkerSkill
- substance-reviewerSkill
- writing-quality-reviewerSkill
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
- Very specific: line 15
Skipped Dimension Handling: If a dimension is skipped (e.g., Dim 3 when no client context), set score: null, skipped: true, skip_reason: …
- Distinctive: line 21
Pass with Warnings: No Dim 1–3 Criticals, but Majors exist in any dimension.
- Distinctive: line 22
Pass: No Critical or Major findings.
- Very specific: line 94
After issue consolidation, scoring, and known-issues matching complete, the main agent updates checkpoint.json:
≈ “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.
Files
scoring-engine Skill
Compute per-dimension review scores, overall grade, and independent release recommendation.
Capabilities
- 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.mdfor detailed criteria
- 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.)
- Release Recommendation (independent safety gate)
- Release Not Recommended: Any Dim 1–3 Critical finding; OR any
Nonexistentcitation on dispositive conclusion - Manual Review Required: Any
Unverifiablecitation 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_recommendationMUST be exactly one of these four string values. Do NOT use free-form text. Userelease_rationalefor 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"
- Release Not Recommended: Any Dim 1–3 Critical finding; OR any
- 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-managerskill
- Merge all findings from Steps 3–5 into
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
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
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.

