Package a reusable CI/CD regression-detection workflow template for downstream repos

[himanshu231204]

Problem

oaeval test and the cicd/ threshold-gate machinery already exist and work, but there's no packaged, drop-in GitHub Actions template showing another repository how to fail its own CI on an evaluation regression. Every adopter currently has to hand-roll this.

Goal

A developer can add regression-gated evaluation to their own repo's CI by copying one file, not by reverse-engineering cicd/plugin.py.

Suggested implementation

Publish either a reusable composite GitHub Action (.github/actions/oaeval-gate/) or a documented workflow snippet (docs/cicd-template.md) that runs oaeval test config.yaml -t <thresholds> and fails the job on threshold violation, with a worked example using the mock provider so it's runnable with zero secrets in a demo.

Files/components

• .github/actions/oaeval-gate/ (proposed new composite action) or docs/cicd-template.md (proposed new doc)
• docs/13_cicd_integration.md (existing — cross-link)

Acceptance criteria

• A copy-pasteable template exists and is demonstrated working (e.g. in this repo's own CI or a linked example repo).

Difficulty

Advanced

---

Filed from a repository audit. Full report: https://claude.ai/code/artifact/134cec04-96c5-4d58-af5e-207def4ab524