feature-ideation: discussion-triggered run fails at workflow setup (zero jobs) — discussion→dispatch bridge

Summary

A feature-ideation run triggered by a new Ideas Discussion fails immediately at workflow setup — zero jobs are created (GitHub: "This run likely failed because of a workflow file issue", conclusion=failure, total_jobs=0). The discussion → workflow_dispatch redispatch bridge (added in #550, building on #457/#448) appears to be broken on the discussion event path, and this looks like its first real-world exercise.

Evidence

Failing run: https://github.com/petry-projects/bmad-bgreat-suite/actions/runs/28534194854 (event=discussion, conclusion=failure, 0 jobs).
Trigger: petry-projects/bmad-bgreat-suite Discussion Org Status — 2026-05-22 #344 "enhanced devops agent persona", category ideas, author don-petry (OWNER, not a bot) → the redispatch job's if: (ideas category + OWNER/MEMBER/COLLABORATOR + non-bot) would have matched. Since no jobs were created at all (not "skipped"), the failure is a workflow-file/validation error at setup, before any job if: is evaluated.
Not a bad copy: bmad's .github/workflows/feature-ideation.yml is byte-identical to the current standards/workflows/feature-ideation.yml template (verbatim enrollment, reusable pinned @e20a8fac # v1).
Never succeeded via discussion: no discussion-triggered feature-ideation run has succeeded anywhere in the org (this path was only recently deployed).

Likely root cause (hypothesis — exact "Invalid workflow file" text not retrievable via API for startup failures)

The caller defines the redispatch inline (redispatch job on discussion, plus an ideate job if: github.event_name != 'discussion' that calls the reusable with a with: block full of ${{ inputs.* }} references). On a discussion event there is no inputs context, and GitHub validates the reusable-call graph (including the ideate job's with:) at setup time regardless of its if: — a plausible trigger for the zero-job compile failure.

Contrast: initiative-planner.yml is a thin caller (uses: + secrets: inherit, no inline redispatch, no with: inputs.*) whose reusable performs the redispatch internally — and it works correctly on discussion events (validated same day: TalkTerm idea:approved → inert epic #333 + DAG).

Impact

Auto-ideation on new Ideas Discussions (feat(feature-ideation): auto-enhance new Ideas Discussions on creation #448/fix(feature-ideation): discussion→dispatch redispatch bridge in the template #550) does not work — every such Discussion trips a failed run.
Fleet-wide latent: any repo running feature-ideation @e20a8fac will hit this the first time a real Ideas Discussion is opened. Blocks enrolling feature-ideation on the stable ring (markets/ContentTwin/google-app-scripts/broodly) until fixed.
Not affected: the schedule/workflow_dispatch ideation paths, and the entire idea→initiative planner/driver core.

Suggested fix directions

Mirror initiative-planner.yml's architecture — make the caller a thin stub and move the discussion→dispatch redispatch into the reusable (which runs under a context where the reusable-call graph validates cleanly), or
Keep the inline bridge but ensure the ideate reusable job's with: does not force ${{ inputs.* }} evaluation on discussion events (e.g. gate via a computed job output / avoid referencing inputs where it's absent).

Please confirm the exact setup error from the run's UI ("Invalid workflow file …") when triaging.

Repro

In a repo with feature-ideation.yml (@e20a8fac), open a Discussion in the Ideas category as an OWNER/MEMBER.
Observe the feature-ideation run fail at setup with 0 jobs.

Refs: #550, #457, #448 · rollout context combsmith/.github-private#978, #515.

Dev-Lead Implementation Plan

Issue: #571 — feature-ideation: discussion-triggered run fails at workflow setup (zero jobs) — discussion→dispatch bridge

Root cause

The feature-ideation.yml caller defines the reusable call (ideate job) with a with: block that references ${{ inputs.* }}. GitHub validates the reusable-call graph (including with:) at setup time regardless of the job if:. On a discussion event there is no inputs context, which trips the zero-job "Invalid workflow file" compile failure — so redispatch never even gets a chance to run.

Scope

Adopt suggested fix direction #2: keep the inline discussion→dispatch bridge, but make the reusable-call with: block never reference ${{ inputs.* }}. Dispatch inputs are computed in a gated prep job and forwarded via needs.prep.outputs.* (an always-valid context). Add a regression lint + test so the anti-pattern can't come back.

Implementation Checklist

standards/workflows/feature-ideation.yml — add a gated prep job that echoes inputs.* to job outputs; change ideate to needs: [prep] and source its with: from needs.prep.outputs.*; keep redispatch bridge
.github/workflows/feature-ideation.yml — apply the same structure to this repo's own caller (dogfood; it carries the same latent bug)
.github/scripts/feature-ideation/lint-caller.sh — fail when a job-level reusable uses: has a with: value referencing ${{ inputs.* }}
.github/workflows/feature-ideation-tests.yml — install pyyaml, add script to shellcheck list, add a "lint caller" gate step
standards/ci-standards.md §9 — document the redispatch-bridge + prep-outputs architecture (currently documents the superseded direct target_discussion pass)

Tests to Write

test/workflows/feature-ideation/lint-caller.bats — clean fixture passes; inputs.*-in-reusable-with: fixture fails with the offending key; step-level action uses: produces no false positive; the real template and real repo caller both pass (regression guards — red before the fix)
fixtures under test/workflows/feature-ideation/fixtures/callers/

Test Command

bats test/workflows/feature-ideation/ (plus bash .github/scripts/feature-ideation/lint-caller.sh <files> and bash .dev-lead/scripts/dev-lead-lint.sh)

Dev-Lead: issue #571 implementation failed — will retry

The engine failed (engine-error — e.g. a timeout or transient engine error). This was attempt 1 of 3; dev-lead will retry automatically. You can also re-apply the dev-lead label to retry now.

Cause: engine-error
Run: https://github.com/petry-projects/.github/actions/runs/28591381864

Dev-Lead: issue #571 implementation failed — will retry

The engine failed (engine-error — e.g. a timeout or transient engine error). This was attempt 2 of 3; dev-lead will retry automatically. You can also re-apply the dev-lead label to retry now.

Cause: engine-error
Run: https://github.com/petry-projects/.github/actions/runs/28604876321

Dev-Lead: cannot implement issue #571 — needs human attention

Automatic retries are exhausted (attempt 3 of 3). The most recent failure was engine-error.

Run: https://github.com/petry-projects/.github/actions/runs/28611239794

Last 40 lines of engine output (redacted)

    [copilot] calling gh copilot (model=openai/o4-mini, timeout=600s, flags=--yolo)
Error: Classic Personal Access Tokens (ghp_) are not supported by Copilot.

The GH_TOKEN environment variable contains a classic PAT.
Please use a Fine-Grained Personal Access Token or another authentication method.

To fix this, you can:
  • Replace the token in GH_TOKEN with a fine-grained PAT
  • Unset GH_TOKEN and run 'gh auth login' to authenticate
  • Unset GH_TOKEN and start 'copilot', then use the '/login' command

Please review the failures above. To retry anyway, remove the dev-lead:needs-human label and re-apply the dev-lead label.

Dev-Lead: cannot implement issue #571 — needs human attention

Automatic retries are exhausted (attempt 4 of 3). The most recent failure was engine-error.

Run: https://github.com/petry-projects/.github/actions/runs/28717149490

Please review the failures above. To retry anyway, remove the dev-lead:needs-human label and re-apply the dev-lead label.

fix(feature-ideation): route caller inputs through a prep job (#571) (#…

Fixed and validated end-to-end ✅

Root cause confirmed. A reusable-workflow call graph (uses: + with:) is evaluated at workflow setup, before and regardless of the calling job's if:. The ideate job's with: referenced ${{ inputs.* }}, and the inputs context isn't populated on the discussion event → the whole run failed at setup with zero jobs, even though ideate was gated if: github.event_name != 'discussion'. (Note: this is a runtime setup failure, not static-lintable — actionlint passes the broken file.)

Fix. Resolve dispatch inputs in a gated prep job and pass them to ideate via needs.prep.outputs.* — an always-valid context that defers the with: evaluation to run time (skipped on discussion). Mirrors the working initiative-planner.yml bridge.

fix(feature-ideation): route caller inputs through a prep job (#571) #615 (this repo) — fixed the canonical template standards/workflows/feature-ideation.yml + this repo's own caller; added lint-caller.sh guard + 10 bats tests wired into feature-ideation-tests.yml; documented the architecture in ci-standards §9.
fix(feature-ideation): route caller inputs through a prep job (#571) .github-private#1113 — applied the same fix to the next-ring dogfood caller + updated its compliance test (which had encoded the old buggy inputs.* contract) and added a feature-ideation: discussion-triggered run fails at workflow setup (zero jobs) — discussion→dispatch bridge #571 guard.

Live validation (petry-projects/.github-private, next ring): opened a real Ideas Discussion and observed the discussion-triggered run:

discussion run → 3 jobs, redispatch: success, prep/ideate correctly skipped — no zero-job startup failure.
redispatch bridged to workflow_dispatch → 4 jobs, prep: success, ideate ran with target_discussion forwarded through needs.prep.outputs.*.

Contrast: the pre-fix discussion runs failed at setup. The bridge now works.

Fleet rollout. The other repos' callers still carry the same latent inputs.* pattern until re-synced to the fixed template — tracked in #614 (each repin/onboard now picks up the fix, and the new lint-caller.sh prevents regressions).

Closing as fixed.

fix(feature-ideation): route caller inputs through a prep job (#571) (#…

Navigation Menu

Repository navigation

feature-ideation: discussion-triggered run fails at workflow setup (zero jobs) — discussion→dispatch bridge #571

Summary

Evidence

Likely root cause (hypothesis — exact "Invalid workflow file" text not retrievable via API for startup failures)

Impact

Suggested fix directions

Repro

Dev-Lead Implementation Plan

Root cause

Scope

Implementation Checklist

Tests to Write

Test Command

Dev-Lead: issue #571 implementation failed — will retry

Dev-Lead: issue #571 implementation failed — will retry

Dev-Lead: cannot implement issue #571 — needs human attention

Dev-Lead: cannot implement issue #571 — needs human attention

1 remaining item

Fixed and validated end-to-end ✅

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

feature-ideation: discussion-triggered run fails at workflow setup (zero jobs) — discussion→dispatch bridge #571

Description

Summary

Evidence

Likely root cause (hypothesis — exact "Invalid workflow file" text not retrievable via API for startup failures)

Impact

Suggested fix directions

Repro

Activity

don-petry commented on Jul 2, 2026

Dev-Lead Implementation Plan

Root cause

Scope

Implementation Checklist

Tests to Write

Test Command

don-petry commented on Jul 2, 2026

Dev-Lead: issue #571 implementation failed — will retry

don-petry commented on Jul 2, 2026

Dev-Lead: issue #571 implementation failed — will retry

don-petry commented on Jul 2, 2026

Dev-Lead: cannot implement issue #571 — needs human attention

don-petry commented on Jul 4, 2026

Dev-Lead: cannot implement issue #571 — needs human attention

1 remaining item

don-petry commented on Jul 4, 2026

Fixed and validated end-to-end ✅

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions