Problem
Between 2026-05-26 10:28 UTC and 10:39 UTC, three agentic-workflow runs in github/gh-aw failed at the GitHub-checkout / repo-permission-check step with the same identity error:
Warning: Repository permission check failed: Sorry. Your account was suspended - https://docs.github.com/rest
##[error]fatal: unable to access 'https://github.com/github/gh-aw/': The requested URL returned error: 403
The process '/usr/bin/git' failed with exit code 128
The 403 was raised before the agent ever started; the post-failure cleanup then failed with Cannot find module '/home/runner/work/_temp/gh-aw/actions/setup_globals.cjs' because the activation step had not yet populated _temp/gh-aw/actions/.
Affected Runs
| Run |
Workflow |
Started (UTC) |
Trigger |
| §26446905995 |
AI Moderator |
10:28:55 |
issue_comment |
| §26447363137 |
Package Specification Extractor |
10:39:07 |
schedule |
| §26447374892 |
Functional Pragmatist |
10:39:24 |
schedule |
Different workflows, different triggers — but the same 403 disposition, so the suspended identity is shared. The 11-minute window plus the fact that subsequent runs succeeded points to a transient suspension or token revocation.
Probable Root Cause
A GitHub App installation token or bot account used by gh-aw's checkout step was suspended for ~11 minutes. Candidates:
- A GitHub App used by gh-aw (e.g., for app-token minting) was rate-limited or temporarily suspended.
- A bot account whose PAT is configured as a workflow secret was suspended.
- The
agentic-workflows-app (or equivalent) installation in github/gh-aw was briefly disabled.
Unknown: whether this was a github/gh-aw-specific event or a broader GitHub-side incident.
Evidence (Log Excerpt — run §26446905995)
Full error context
2026-05-26T10:41:03.9163955Z ##[warning]Repository permission check failed: Sorry. Your account was suspended - https://docs.github.com/rest
2026-05-26T10:41:03.9176907Z ##[warning]⚠️ Repository permission check failed: Sorry. Your account was suspended - https://docs.github.com/rest
...
2026-05-26T10:42:01.2065615Z ##[error]fatal: unable to access 'https://github.com/github/gh-aw/': The requested URL returned error: 403
2026-05-26T10:42:01.2073943Z The process '/usr/bin/git' failed with exit code 128
2026-05-26T10:42:16.7978174Z ##[error]fatal: unable to access 'https://github.com/github/gh-aw/': The requested URL returned error: 403
2026-05-26T10:42:16.7996275Z The process '/usr/bin/git' failed with exit code 128
2026-05-26T10:42:31.8815925Z ##[error]fatal: unable to access 'https://github.com/github/gh-aw/': The requested URL returned error: 403
2026-05-26T10:42:32.0004095Z ##[error]The process '/usr/bin/git' failed with exit code 128
Proposed Remediation
- Identify the suspended principal. Cross-reference the failing runs' OIDC subject / app installation ID against repository app installation logs to determine which identity was suspended.
- Improve diagnostic. The current pre-activation warning is logged but the failure surfaces three steps later as a generic
git 128. Surface the upstream warning into the failure-issue body so the next human triager sees account suspended immediately.
- Resilient cleanup. Three runs additionally crashed in the failure-handler path with
Cannot find module 'setup_globals.cjs' because the pre-activation download did not complete. The conclusion step should not require artifacts the activation step is responsible for creating; it should fail-soft when those are missing.
Success Criteria
- The suspended principal is identified and the 11-minute outage explained (post-mortem comment on this issue is sufficient).
- The failure-handler path produces a usable failure-issue body even when activation never populated
_temp/gh-aw/actions/ — no Cannot find module 'setup_globals.cjs' traceback.
- (Optional) The pre-activation
Repository permission check failed warning is surfaced in the auto-generated [aw] <workflow> failed issue body when present.
Verification
- After fixes: simulate an account-suspended-style 403 in a synthetic run (e.g., revoke the token mid-run via a test app) and confirm:
- The failure-issue body contains the suspension warning text, not just
exit code 128.
- No
Cannot find module 'setup_globals.cjs' error appears in the conclusion log.
References
Generated by 🔍 [aw] Failure Investigator (6h) · opus47 18.6M · ◷
Problem
Between 2026-05-26 10:28 UTC and 10:39 UTC, three agentic-workflow runs in
github/gh-awfailed at the GitHub-checkout / repo-permission-check step with the same identity error:The 403 was raised before the agent ever started; the post-failure cleanup then failed with
Cannot find module '/home/runner/work/_temp/gh-aw/actions/setup_globals.cjs'because the activation step had not yet populated_temp/gh-aw/actions/.Affected Runs
issue_commentschedulescheduleDifferent workflows, different triggers — but the same 403 disposition, so the suspended identity is shared. The 11-minute window plus the fact that subsequent runs succeeded points to a transient suspension or token revocation.
Probable Root Cause
A GitHub App installation token or bot account used by gh-aw's checkout step was suspended for ~11 minutes. Candidates:
agentic-workflows-app(or equivalent) installation ingithub/gh-awwas briefly disabled.Unknown: whether this was a
github/gh-aw-specific event or a broader GitHub-side incident.Evidence (Log Excerpt — run §26446905995)
Full error context
Proposed Remediation
git 128. Surface the upstream warning into the failure-issue body so the next human triager seesaccount suspendedimmediately.Cannot find module 'setup_globals.cjs'because the pre-activation download did not complete. The conclusion step should not require artifacts the activation step is responsible for creating; it should fail-soft when those are missing.Success Criteria
_temp/gh-aw/actions/— noCannot find module 'setup_globals.cjs'traceback.Repository permission check failedwarning is surfaced in the auto-generated[aw] <workflow> failedissue body when present.Verification
exit code 128.Cannot find module 'setup_globals.cjs'error appears in the conclusion log.References
Related to [aw-failures] Failure Investigator (6h) — 14 failures, 3 clusters (2026-05-26 08:00–14:00 UTC) #34938