Smoke Copilot activation fails when repository checkout removes gh-aw setup action before cleanup
Los mantenedores suelen responder en 1 día
@lpcox ya está trabajando en esto.
Desde el 1/11/2026.
- #15859 de @copilot-swe-agent — cerrado sin fusionar
Evaluación
Este issue todavía no se ha evaluado.
Descripción
Problem
The Smoke Copilot workflow fails during activation's post-job cleanup because a later repository checkout deletes the local gh-aw setup action. The agent never starts, so the smoke test does not exercise Copilot or the MCP gateway.
Observed failure
- Run: https://github.com/github/gh-aw-mcpg/actions/runs/41350798789
- Failed job: https://github.com/github/gh-aw-mcpg/actions/runs/41350798789/job/124980311169
- Trigger:
workflow_dispatchonmain - Revision:
95567d9616528b160a0716bd201e3eced82ea095 - Workflow:
.github/workflows/smoke-copilot.lock.yml - Failed step: Post Setup Scripts
Can't find 'action.yml', 'action.yaml' or 'Dockerfile' under '/home/runner/work/gh-aw-mcpg/gh-aw-mcpg/actions/setup'. Did you forget to run actions/checkout before running your local action?
All regular activation steps succeeded, including prompt preparation and artifact uploads. The cleanup failure marked activation as failed, and the agent, conclusion, and safe_outputs jobs were skipped.
Root cause
The generated workflow uses the same workspace directory for two different repositories:
-
Checkout actions folder checks out
github/gh-awinto the workspace root, with a sparse checkout ofactions. -
Setup Scripts invokes
uses: ./actions/setup. Runtime scripts are copied to${{ runner.temp }}/gh-aw/actions, but the local action definition remains in the workspace. -
Checkout .github and .agents folders checks out
github/gh-aw-mcpginto that same workspace root. The run log explicitly reports:2026-11-01T21:13:09.7718579Z Deleting the contents of '/home/runner/work/gh-aw-mcpg/gh-aw-mcpg' -
At job teardown, GitHub Actions attempts to load the local setup action for its post step, but
actions/setup/action.ymlis gone.
The second checkout includes actions/setup in its sparse-checkout list, but that selects paths from github/gh-aw-mcpg; it cannot preserve files from the previous github/gh-aw checkout. Copying runtime scripts to the runner temp directory likewise does not preserve the local action metadata that the runner needs for cleanup.
Proposed fix
Fix the gh-aw workflow generator so the setup action checkout survives every later checkout until all post-job hooks finish. Do not hand-edit this generated lock file as the permanent fix.
Prefer an isolated layout: check out github/gh-aw into a dedicated workspace-relative directory, update local setup-action references accordingly, and ensure subsequent target-repository checkouts cannot delete that directory. This requires coordinating both checkout locations; merely adding path: to the first checkout while continuing to replace the workspace root is insufficient. Update any script paths that depend on the layout.
An alternative worth evaluating is invoking the setup action directly through a pinned github/gh-aw/actions/setup@<revision> reference, which lets the runner manage its action directory separately from the target checkout. Confirm compatibility with setup's repository-relative dependencies before choosing this approach.
Once the generator fix is available, regenerate the affected workflows in this repository using the supported compiler path. Inspect other generated workflows for the same checkout/setup pattern; recurrence beyond this run has not yet been established.
Acceptance criteria
- The setup action definition and any post-step dependencies remain available through job teardown after the target repository checkout.
- Add a generator regression test covering setup from
github/gh-awfollowed by a checkout of a different repository. - Verify in a hosted Actions run that Post Setup Scripts succeeds and the Smoke Copilot
agentjob starts when activation otherwise succeeds. - Regenerated locks preserve the intended repository checkout, prompt preparation, and setup/cleanup behavior.
Scope of evidence
This diagnosis is based on the failed run's job statuses and logs and the workflow configuration at the failed revision. No patch or rerun has been performed.
- Lenguaje dominante
- Go
- Estrellas
- 184
- Forks
- 58
- Merge medio
- 7 h 44 min
- PR fusionados (30 d)
- 186
Preparar el entorno
Inicia el contenedor de desarrollo del proyecto en tu navegador, con tu propia cuenta de GitHub.
- Incluye un Dockerfile o un archivo de Docker Compose
- Sin plantilla de pull request
- Leer la guía de contribución
Primeros pasos
- Lee el issue completo y luego la guía de contribución del proyecto.
- Comenta en el issue que vas a ocuparte — evita que dos personas hagan lo mismo.
- Haz un fork del repositorio y trabaja en una rama.
- Abre un pull request que haga referencia al número del issue.
Más de github/gh-aw-mcpg
-
agentic-workflows maintenance
Dificultad 2/5 1-3 horas Aptitud para principiantes 72/100
github/gh-aw-mcpg#16043 · 1 comentario ·
Los mantenedores suelen responder en 1 día
-
Smoke AllowOnly - Issue GroupAbierto
Dificultad 5/5 Más de una semana Aptitud para principiantes 1/100
github/gh-aw-mcpg#16045 ·
Los mantenedores suelen responder en 1 día
-
Smoke AllowOnly: 42121621134Abierto
Dificultad 4/5 3-5 días Aptitud para principiantes 35/100
github/gh-aw-mcpg#16044 ·
Los mantenedores suelen responder en 1 día
-
automation repo-assist
Dificultad 3/5 Medio día Aptitud para principiantes 5/100
github/gh-aw-mcpg#15847 ·
Los mantenedores suelen responder en 1 día
-
automation mcp-gateway test
Dificultad 1/5 Menos de una hora Aptitud para principiantes 5/100
github/gh-aw-mcpg#6522 ·
Los mantenedores suelen responder en 1 día
Todos los issues de github/gh-aw-mcpg
Issues similares
-
area/docs framework/hugo help wanted language/markdown
Dificultad 1/5 Menos de una hora Aptitud para principiantes 88/100
Los mantenedores suelen responder en 1 día
-
clawsweeper:fix-shape-clear clawsweeper:queueable-fix clawsweeper:source-repro impact:other issue-rating: 🦞 diamond lobster no-stale P2
Dificultad 2/5 1-3 horas Aptitud para principiantes 75/100
openclaw/crabbox#3060 · 1 comentario · 1 reacción ·
Los mantenedores suelen responder en 1 día
-
Dificultad 2/5 1-3 horas Aptitud para principiantes 76/100
kubeflow/semantic-operator#52 ·
Los mantenedores suelen responder en 10 días
-
Dificultad 2/5 1-3 horas Aptitud para principiantes 78/100
Los mantenedores suelen responder en 1 día
-
enhancement help wanted
Dificultad 2/5 1-3 horas Aptitud para principiantes 78/100
PramodBasavarajMenasi/siftwise#52 ·
Los mantenedores suelen responder en 1 día