Require explicit terminal safe-output in Daily Agent of the Day Blog Writer

[Copilot]

The Daily Agent of the Day Blog Writer run could complete with success while emitting no safe outputs, which triggers failure reporting despite no job-level errors. The workflow prompt had a no-action guideline, but it did not enforce a terminal safe-output decision in blocked/incomplete paths.

• Prompt contract hardening

  • Added a new Mandatory safe-output completion section to .github/workflows/daily-agent-of-the-day-blog-writer.md.
  • Requires the agent to finish with exactly one terminal safe-output:
    • create_pull_request when content is produced and ready for PR.
    • noop when analysis is complete and no action is needed.
    • report_incomplete when execution is blocked by infra/tooling constraints (e.g., permission-denied, unavailable MCP/tool access).
  • Explicitly prohibits ending with plain text and no safe-output call.

• Compiled workflow update

  • Regenerated .github/workflows/daily-agent-of-the-day-blog-writer.lock.yml so runtime behavior reflects the prompt change.

### 8) Mandatory safe-output completion

You **MUST** finish by calling exactly one safe-output tool:

- `create_pull_request` when you created the blog post and are ready to open the PR.
- `noop` when no action is needed after valid analysis.
- `report_incomplete` when blocked by infrastructure/tooling failures.

Never end with plain text only and no safe-output call.