Summary
The agent job in workflows using gh-aw v0.77.5 on github/actions-sales (Actions Specialist Intake Triage) fails immediately at the Execute GitHub Copilot CLI step with:
Error: Cannot find module '/home/runner/work/_temp/gh-aw/actions/merge_awf_model_multipliers.cjs'
at Function._resolveFilename (node:internal/modules/cjs/loader:1430:15)
...
Node.js v22.23.1
code: 'MODULE_NOT_FOUND'
##[error]Process completed with exit code 1.
The activation job succeeds. The agent job blows up before Copilot ever runs. Failure signature is identical across all 5 runs, so this reads like a v0.77.5 setup step that's not materializing the .cjs shim into _temp/gh-aw/actions/.
Recurrence pattern (last 8 days)
All are the same intake-triage workflow, same signature, same error, always the agent job:
| Issue |
Failed run |
Date |
| github/actions-sales#924 (latest run) |
28559150351 |
2026-07-02 01:33 UTC |
| github/actions-sales#924 |
(prior in same issue) |
2026-07-01 19:38 UTC |
| github/actions-sales#922 |
|
2026-06-29 |
| github/actions-sales#919 |
|
2026-06-26 |
| github/actions-sales#914 |
|
2026-06-25 |
Blast radius: every specialist intake through github/actions-sales on those days silently missed AI triage. Humans have to manually route.
Environment (from failed run 28559150351)
GH_AW_VERSION: v0.77.5
COPILOT_MODEL: claude-sonnet-4.6
GH_AW_WORKFLOW_ID_SANITIZED: intaketriage
COPILOT_AGENT_RUNNER_TYPE: STANDALONE
- Node.js v22.23.1
- Runs on ubuntu-latest (github-hosted)
Suggested fast path
Check what step is supposed to write merge_awf_model_multipliers.cjs into _temp/gh-aw/actions/ in v0.77.5. Likely one of the pre-agent compile/materialize steps is either skipped, conditioned wrong, or the shim was renamed/moved and the loader wasn't updated.
Context
Filing on behalf of @austenstone (specialist queue owner) — 5 silent misses is over the threshold for opening the bug upstream rather than continuing to comment on symptomatic actions-sales issues. Happy to test a candidate fix on our intake workflow if a v0.77.6 lands.
Summary
The
agentjob in workflows usinggh-aw v0.77.5ongithub/actions-sales(Actions Specialist Intake Triage) fails immediately at the Execute GitHub Copilot CLI step with:The
activationjob succeeds. Theagentjob blows up before Copilot ever runs. Failure signature is identical across all 5 runs, so this reads like a v0.77.5 setup step that's not materializing the.cjsshim into_temp/gh-aw/actions/.Recurrence pattern (last 8 days)
All are the same intake-triage workflow, same signature, same error, always the
agentjob:Blast radius: every specialist intake through
github/actions-saleson those days silently missed AI triage. Humans have to manually route.Environment (from failed run 28559150351)
GH_AW_VERSION: v0.77.5COPILOT_MODEL: claude-sonnet-4.6GH_AW_WORKFLOW_ID_SANITIZED: intaketriageCOPILOT_AGENT_RUNNER_TYPE: STANDALONESuggested fast path
Check what step is supposed to write
merge_awf_model_multipliers.cjsinto_temp/gh-aw/actions/in v0.77.5. Likely one of the pre-agent compile/materialize steps is either skipped, conditioned wrong, or the shim was renamed/moved and the loader wasn't updated.Context
Filing on behalf of @austenstone (specialist queue owner) — 5 silent misses is over the threshold for opening the bug upstream rather than continuing to comment on symptomatic
actions-salesissues. Happy to test a candidate fix on our intake workflow if a v0.77.6 lands.