Behavior
Per-trait output
Each changed file is scored against every trait on the selected model. The PR comment lists the per-trait scores, tier labels, headroom, and confidence for each file.
Deterministic
Same content, same model, same score across runs. Scores are reproducible for audit and comparison.
Model selection
The model is configured per workflow. Any public model handle is accepted.
Threshold flagging
An optional composite threshold flags files whose composite score falls below it, leaving the pass/fail decision to a downstream check.
Setup
Add this workflow to .github/workflows/u22a8.yml:
name: Score docs
on:
pull_request:
paths: ["**/*.md"]
permissions:
pull-requests: write
jobs:
score:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- uses: u22a8/plugins/actions/score-docs@main
with:
profile: u22a8.technical-writing
threshold: 60
Options
profile
Handle of the model to score against. Defaults to u22a8.compelling-readme.
threshold
Optional 0–100 composite score threshold. Files below are flagged in the PR comment.
files
Glob pattern for files to score. Defaults to **/*.md.