Skip to content

ci: guard Testbox workflows behind repository check (missed in #48)#49

Merged
LightDriverCS merged 2 commits into
mainfrom
fork/guard-testbox-workflows
Jun 3, 2026
Merged

ci: guard Testbox workflows behind repository check (missed in #48)#49
LightDriverCS merged 2 commits into
mainfrom
fork/guard-testbox-workflows

Conversation

@LightDriverCS

Copy link
Copy Markdown

Completes #48: the four Testbox experiment workflows were missed in the first sweep — two of them ("Blacksmith Testbox", "Blacksmith Build Artifacts Testbox") sat queued for 30+ minutes on #48 itself, demonstrating the disease. Same one-line guard per job; functional CI untouched.

Real behavior proof

Behavior or issue addressed: Four Testbox workflows queue forever on this fork (Blacksmith pools incl. 32vcpu not provisioned), leaving eternal pending checks on every PR.

Real environment tested: This repository's live Actions history — the two Testbox checks observed stuck-queued for 30+ minutes on PR #48 today (runs 26906163699 / 26906163598, status queued, no conclusion).

Exact steps or command run after this patch: ruby YAML parse of all 4 edited files; grep verification of exactly one job-level if per file.

Evidence after fix: Copied output:

$ gh run view 26906163699 --json status,conclusion  # before guards
Blacksmith Build Artifacts Testbox  queued  -
$ for f in ...; do ruby -ryaml -e "YAML.load_file(...)" && echo ok; done
✓ all 4 parse, exactly 1 job-level if each

Observed result after fix: On BenchAGI/openclaw these jobs evaluate if to false and report skipped; upstream behavior unchanged (condition preserved verbatim).

What was not tested: Upstream execution (this fork cannot exercise upstream CI; guard preserves their conditions verbatim).

🤖 Generated with Claude Code

Cory Shelton and others added 2 commits June 3, 2026 13:28
Same fork-guard as #48 for the four Testbox experiment workflows — their
Blacksmith pools (incl. 32vcpu) queue forever on this fork; two of them sat
'pending' on #48 itself for 30+ minutes.

Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>
…izmor template-injection)

Canonical zizmor remediation — the expansion moves to env: and the
PowerShell block reads $env:TARGET_REF. Surfaced because #49 touched the
file, making security-fast's changed-files zizmor audit it for the first
time; the finding predates the guards.
@LightDriverCS
LightDriverCS merged commit cf3be6b into main Jun 3, 2026
76 of 77 checks passed
@LightDriverCS

Copy link
Copy Markdown
Author

Anvil review: fixed-ready.

  • Diff is workflow-only: four Testbox workflows are gated to openclaw/openclaw; the Windows probe also routes target_ref through step env so zizmor no longer flags template injection in the PowerShell block.
  • Targeted local proof: git diff --check; Ruby YAML parse for all 4 changed workflows; pre-commit zizmor on the 4 workflows with Python 3.14; pre-commit actionlint on the 4 workflows.
  • Live proof on 8fc13af75f8cd0372c6e2055ea9157ab00ea5ab3: fork Testbox workflows skipped as intended; security-fast and workflow sanity passed; full CI is green after rerunning check-lint attempt 2. Attempt 1 was a runner shutdown after oxlint had already reported 0 warnings and 0 errors.
  • Firebase/App Hosting: not touched.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant