agentleFS
solo-istsolo-ist/prose/.agents/skills/pipeline-scorer/SKILL.md

pipeline-scorer

Skill8 starsChanged 5 months agoMITView on GitHub

Score PR review findings using a quantitative rubric. Outputs a 1-10 composite score across 6 dimensions to inform auto-fix vs human-review routing.

The file

We split this skill into the rules it gives an agent

1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

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

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

    Don't double-count: If the review lists 5 instances of the same pattern (e.g., missing validatePath() in 5 handlers), that's one concern …

    line 59

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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

Pipeline Scorer

Quantitative assessment of PR review findings to drive triage routing.

Usage

/pipeline-scorer <pr-number>

Rubric

Scores each dimension 1-5, then computes a composite score (average x 2, range 1-10):

Dimension1 (Low)5 (High)
Scope1-2 files changedCross-cutting changes
Severity mixStyle/nitpicks onlySecurity/data-loss issues
Change typeConfig/docsCore architecture
Test impactNo test impactUntested critical paths
API surfaceInternal onlyBreaking public API
Dependency riskNo dep changesMajor dep overhaul

Routing Thresholds

ScoreRouteWhat it means downstream
1-3auto-fixIssues are trivial — pipeline-fix.yml pushes a fix commit
4-6reviewNeeds human verification — may become auto-fix-verify or hitl-full depending on PE risk
7-10complexFull human review required — always routes hitl-full

Note: the scorer's threshold label is an input to the orchestrator, which applies its own routing matrix combining it with the PE risk level. The scorer does not directly dispatch anything.

Calibration Examples

Score ~2: Style-only fix

  • Review found: two trailing whitespace warnings, one unused import
  • Scope: 1 file | Severity: 1 | Type: 1 | Test: 1 | API: 1 | Deps: 1
  • Composite: 2.0 → auto-fix

Score ~4: Isolated bug fix

  • Review found: off-by-one in pagination, missing null check in API handler
  • Scope: 2 | Severity: 3 | Type: 2 | Test: 3 | API: 1 | Deps: 1
  • Composite: 4.0 → review

Score ~8: Security fix touching core architecture

  • Review found: SQL injection in query builder, missing input validation on 3 API endpoints
  • Scope: 4 | Severity: 5 | Type: 4 | Test: 4 | API: 3 | Deps: 1
  • Composite: 7.0 → complex

Common Pitfalls

  • Sparse reviews inflate severity. A review with one nit and nothing else should score low (1-2), not medium. The absence of findings is signal.
  • Don't double-count. If the review lists 5 instances of the same pattern (e.g., missing validatePath() in 5 handlers), that's one concern with wide scope — not 5 high-severity findings.
  • Config changes are rarely high-severity. Changing tsconfig.json, package.json scripts, or CI workflow YAML is almost always scope=1, severity=1 unless it introduces a new dependency or changes build behavior.
  • Test impact != test changes. A PR that changes core logic with no corresponding test updates should score high on test_impact even if the PR doesn't touch test files.

Workflow

  1. Fetch the code review comment from the PR
  2. Send to Codex Sonnet with the scoring rubric
  3. Parse dimension scores and compute composite
  4. Output structured analysis with sentinel for downstream pipeline

Output Format

The output includes a machine-readable sentinel: <!-- scorer-output: {"score":N,"dimensions":{...},"threshold":"..."} -->

And a human-readable breakdown table with rationale for each dimension score.

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.