mirror of
https://github.com/NousResearch/hermes-agent.git
synced 2026-07-21 16:18:55 +00:00
#66373 swapped GITHUB_TOKEN -> AUTOFIX_BOT_PAT across the workflows and #66577 restored the `|| github.token` fork fallback for detect-changes and the label gates -- but it missed the ci-timings "Collect timings and generate report" step, which still passes a bare AUTOFIX_BOT_PAT. On fork PRs that PAT is empty, so timings_report.py hard-fails at expect_env("GITHUB_TOKEN") before it can reach its own "degraded run must never redden the PR" soft-fail path. Every fork PR gets a red run from this advisory job (e.g. #66573). - ci.yml: apply the same `secrets.AUTOFIX_BOT_PAT || github.token` fallback to the timings step. github.token has `actions: read`, enough to read the run's job/step durations on forks. - timings_report.py: treat a missing/empty GITHUB_TOKEN as a degraded run (TimingsUnavailable) instead of a hard ValueError, so this whole class of failure can never redden a PR again even if a future workflow drops the token. Still writes no JSON, so no empty baseline is ever cached. |
||
|---|---|---|
| .. | ||
| ci.yml | ||
| contributor-check.yml | ||
| deploy-site.yml | ||
| docker-lint.yml | ||
| docker.yml | ||
| docs-site-checks.yml | ||
| history-check.yml | ||
| js-autofix.yml | ||
| js-tests.yml | ||
| lint.yml | ||
| lockfile-diff.yml | ||
| osv-scanner.yml | ||
| skills-index-freshness.yml | ||
| skills-index.yml | ||
| supply-chain-audit.yml | ||
| tests.yml | ||
| upload_to_pypi.yml | ||
| uv-lockfile-check.yml | ||