-
Notifications
You must be signed in to change notification settings - Fork 295
Description
Problem
The Contribution Check workflow (contribution-check) has been failing consistently since 2026-03-15 due to a safe outputs processing error. The agent generates an add_labels safe output call that uses a temporary_id (aw_report1) as the item_number, but the add_labels handler does not support temporary ID resolution.
Error Message
##[warning]Invalid item number: aw_report1
##[error]✗ Message 6 (add_labels) failed: Invalid item number: aw_report1
##[error]1 safe output(s) failed:
- add_labels: Invalid item number: aw_report1
Failure Pattern
- Frequency: 15+ consecutive failures on 2026-03-15 (multiple runs/day)
- Last success: Run Network Permissions Feature Test Results - MCP Container Security Validation #164 on 2026-03-14
- Failure rate: 28/30 recent runs
- Failure step:
safe_outputsjob → "Process Safe Outputs" step
Root Cause
The Copilot agent produces this sequence of safe output calls:
create_issuewithtemporary_id: aw_report1→ creates issue (succeeds ✅)add_labelswithitem_number: aw_report1→ fails becauseadd_labelshandler doesn't resolve temporary IDs ❌
The add_labels safe output handler resolves the temporary ID map only after all messages are processed, or the resolution mechanism is not implemented for add_labels. Other handlers (add_comment) do support temporary ID references.
Evidence
- Run §23104133000 — safe_outputs job #67110634450
- The
create_issue(message 5) succeeds and registersaw_report1 → github/gh-aw#21029 - The
add_labels(message 6) immediately fails:Invalid item number: aw_report1
Fix Options
Option 1 (Recommended — fix safe outputs handler): Implement temporary ID resolution in the add_labels handler, consistent with add_comment behavior.
Option 2 (Workaround — fix agent prompt): Update the Contribution Check workflow prompt to instruct the agent not to use temporary IDs in add_labels calls — instead, use the actual issue number once created.
Option 3 (Workaround — reorder calls): Move the add_labels call after the create_issue in the agent's output so that a reference can be made by absolute issue number. This isn't fully reliable since the agent's output order varies.
Impact
- Contribution Check workflow fails on every run where the agent creates a new issue report
- The agent itself runs successfully — the PR reviews and comments are made — but the workflow conclusion is
failure - Creates unnecessary noise in CI failure alerts
References
- Run: §23104133000
- Failing job: https://github.com/github/gh-aw/actions/runs/23104133000/job/67110634450
- Detected by: §23105821296 (Workflow Health Manager)
Related to #19352
Generated by Workflow Health Manager - Meta-Orchestrator · ◷
- expires on Mar 16, 2026, 7:36 AM UTC