Debug mode

GrillMyCode supports GitHub Actions' built-in debug logging mechanism. When enabled, the action emits extra log entries that show exactly what was resolved and what was sent to the AI provider — useful for troubleshooting unexpected assessment results or verifying that context inputs are being picked up correctly.

What gets logged

Debug entry	Contents
Resolved inputs	Every parsed input value after defaults and clamping have been applied. Sensitive fields (github_token, api_key) are shown as [REDACTED].
Code after comment stripping	The full code content after comments have been removed. Omitted when keep_comments: true — a message is logged instead confirming no comments were removed.
Prompt messages	The full [system, user] message array passed to the AI provider, including all injected code content, file lists, assignment context, and instructor instructions.
PDF: launching Chromium	The resolved PUPPETEER_EXECUTABLE_PATH used to launch the headless browser for PDF generation.
PDF: md-to-pdf returned	The content type, byte length, and Buffer status of the PDF returned by md-to-pdf — useful for diagnosing PDF generation failures.
PDF upload	The filename, byte count, and upload URL used when attaching the PDF to the gmc-assessments release.

How to enable it

Set the ACTIONS_STEP_DEBUG secret (or variable) to true in your repository or organization settings.

1. Go to Settings → Secrets and variables → Actions
2. Under Secrets, add a new secret:
   • Name: ACTIONS_STEP_DEBUG
   • Value: true

Debug output will appear in the workflow run logs under the step that runs this action. Lines are prefixed with ##[debug].

tip

Setting ACTIONS_STEP_DEBUG as a secret rather than a variable prevents the value from being visible in workflow run summaries. GitHub recommends this for debug flags that you enable temporarily during troubleshooting.

Re-running with debug enabled

You can also enable debug output for a single re-run without adding a permanent secret:

1. Open the failed (or any past) workflow run
2. Click Re-run jobs → Re-run all jobs
3. Check Enable debug logging in the dialog

This enables ACTIONS_STEP_DEBUG only for that run, with no permanent repository changes required.

What it does not log

• The raw AI response. When the instructor repository is configured, it is saved there as raw-ai-output.md on every run, debug mode or not
• Git diff content (already visible in core.info output under normal logging)
• Code content before comment stripping