Problem & Motivation
When a custom-code repo is transferred to a client's Vercel team, deploys keep firing and fail with a generic error. This (a) generates constant noise and (b) blocks any useful org-wide alert on consecutive Vercel-deploy failures. Recent postmortem: the noise conflated two distinct issues during initial diagnosis and low weekend merge activity hid the real failure for ~48h.
Scope & Decomposition
- Add a
transferred input (default false) to the deploy.yml reusable in platform. When true, the deploy job logs a skip message and exits 0 without invoking Vercel.
- Update the callers of the two transferred repos (
umh-custom-code, profile-behavior-custom-code) to pass transferred: true.
- Document the pattern in
platform/README.md (when to use, how to set the flag, how to verify).
Note: org-wide alert on consecutive Vercel-deploy failures will be tracked in a separate issue, blocked by this one (the alert is only useful once the flag eliminates transferred-repo noise).
Acceptance Criteria
Risks & Constraints
- If the flag is set by accident on an active repo, deploys silently succeed (skipped) without error — the skip log message must be visible and grep-able.
- The reusable changes its contract (new input) — all callers pinned to
@v1 will see it, so the input MUST have default false to avoid breaking anyone.
Edge Cases
- Transferred repo whose caller still runs
ci jobs — the flag must affect only the deploy, not CI (lint/typecheck/test/build remain useful if someone is still maintaining the repo).
- "Semi-transferred" repo where we still want CI but not deploy — covered by the previous point.
Sub-issues
Work each sub-issue on its own branch, opening PRs against the epic branch. Merge the epic branch into main when all sub-issues are complete.
Problem & Motivation
When a custom-code repo is transferred to a client's Vercel team, deploys keep firing and fail with a generic error. This (a) generates constant noise and (b) blocks any useful org-wide alert on consecutive Vercel-deploy failures. Recent postmortem: the noise conflated two distinct issues during initial diagnosis and low weekend merge activity hid the real failure for ~48h.
Scope & Decomposition
transferredinput (defaultfalse) to thedeploy.ymlreusable inplatform. Whentrue, the deploy job logs a skip message and exits 0 without invoking Vercel.umh-custom-code,profile-behavior-custom-code) to passtransferred: true.platform/README.md(when to use, how to set the flag, how to verify).Note: org-wide alert on consecutive Vercel-deploy failures will be tracked in a separate issue, blocked by this one (the alert is only useful once the flag eliminates transferred-repo noise).
Acceptance Criteria
deploy.ymlreusable accepts atransferredinput (defaultfalse); whentrue, the deploy job logs a skip message and exits 0 without invoking Vercel.umh-custom-codeandprofile-behavior-custom-codecallers passtransferred: trueand stop producing failed runs.platform/README.md(when to use, how to set the flag, how to verify).Risks & Constraints
@v1will see it, so the input MUST have defaultfalseto avoid breaking anyone.Edge Cases
cijobs — the flag must affect only the deploy, not CI (lint/typecheck/test/build remain useful if someone is still maintaining the repo).Sub-issues
transferredinput todeploy.ymlreusable #21 � Addtransferredinput todeploy.ymlreusabletransferred: true#22 � Update transferred callers (umh-custom-code, profile-behavior-custom-code) to passtransferred: trueplatform/README.md#23 � Document the transferred-repo pattern inplatform/README.md