mathias/doodle, generated from this template on 2026-10-08, has had a red main since its first commit: 15 of 16 workflow runs are failure, the only success being a tag push that skips deploy. Its task check passes locally and in a fresh clone.
The template's cd.yml runs build and deploy whenever the guard says "not a template". deploy patches k3s/apps/<name>/deployment.yaml in mathias/infra. A new repo has no such manifest: ADR-0001 in mathias/local-dev records the infra lane as explicitly unowned, so it is created later, by hand.
Hypothesis (unconfirmed)
Every repo generated from this template is red from commit 1 until its infra manifest exists, and nothing in the template says so. Not confirmed: the Gitea MCP tools return only a run's conclusion, not which job failed or its log. See mathias/doodle#3, whose first acceptance item identifies the failing job.
If confirmed, the effect is wider than one repo: red-from-birth teaches people that red is normal in a new repo, and it makes clause 6 of the agent-ready contract unsatisfiable for every new project until a human does the infra step.
Acceptance Criteria
The failing job and step for mathias/doodle run 9739 are identified from the job log and posted here (or the hypothesis is refuted and this issue is closed with the real cause).
The same is checked on one other generated repo that has no infra manifest yet, if one exists, so the finding is not generalised from a single repo. If none exists, say so.
If confirmed: a follow-up issue is filed describing the template change (for example deploy skipping with an explicit "no manifest in infra" message instead of failing). No template change is made under this issue.
## Observation
`mathias/doodle`, generated from this template on 2026-10-08, has had a red `main` since its first commit: 15 of 16 workflow runs are `failure`, the only `success` being a tag push that skips `deploy`. Its `task check` passes locally and in a fresh clone.
The template's `cd.yml` runs `build` and `deploy` whenever the guard says "not a template". `deploy` patches `k3s/apps/<name>/deployment.yaml` in `mathias/infra`. A new repo has no such manifest: ADR-0001 in `mathias/local-dev` records the infra lane as explicitly unowned, so it is created later, by hand.
## Hypothesis (unconfirmed)
Every repo generated from this template is red from commit 1 until its infra manifest exists, and nothing in the template says so. **Not confirmed:** the Gitea MCP tools return only a run's conclusion, not which job failed or its log. See `mathias/doodle#3`, whose first acceptance item identifies the failing job.
If confirmed, the effect is wider than one repo: red-from-birth teaches people that red is normal in a new repo, and it makes clause 6 of the agent-ready contract unsatisfiable for every new project until a human does the infra step.
## Acceptance Criteria
- [ ] The failing job and step for `mathias/doodle` run 9739 are identified from the job log and posted here (or the hypothesis is refuted and this issue is closed with the real cause).
- [ ] The same is checked on one other generated repo that has no infra manifest yet, if one exists, so the finding is not generalised from a single repo. If none exists, say so.
- [ ] If confirmed: a follow-up issue is filed describing the template change (for example `deploy` skipping with an explicit "no manifest in infra" message instead of failing). No template change is made under this issue.
**Risk:** LOW
## Source
`mathias/doodle` retro, `docs/retro/2026-10-08-remotion-bootstrap.md`; `mathias/doodle#3`.
---
_Created via git-mcp on behalf of @mathias_
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Observation
mathias/doodle, generated from this template on 2026-10-08, has had a redmainsince its first commit: 15 of 16 workflow runs arefailure, the onlysuccessbeing a tag push that skipsdeploy. Itstask checkpasses locally and in a fresh clone.The template's
cd.ymlrunsbuildanddeploywhenever the guard says "not a template".deploypatchesk3s/apps/<name>/deployment.yamlinmathias/infra. A new repo has no such manifest: ADR-0001 inmathias/local-devrecords the infra lane as explicitly unowned, so it is created later, by hand.Hypothesis (unconfirmed)
Every repo generated from this template is red from commit 1 until its infra manifest exists, and nothing in the template says so. Not confirmed: the Gitea MCP tools return only a run's conclusion, not which job failed or its log. See
mathias/doodle#3, whose first acceptance item identifies the failing job.If confirmed, the effect is wider than one repo: red-from-birth teaches people that red is normal in a new repo, and it makes clause 6 of the agent-ready contract unsatisfiable for every new project until a human does the infra step.
Acceptance Criteria
mathias/doodlerun 9739 are identified from the job log and posted here (or the hypothesis is refuted and this issue is closed with the real cause).deployskipping with an explicit "no manifest in infra" message instead of failing). No template change is made under this issue.Risk: LOW
Source
mathias/doodleretro,docs/retro/2026-10-08-remotion-bootstrap.md;mathias/doodle#3.Created via git-mcp on behalf of @mathias