windmill/docs/autonomous-mode.md

Autonomous Mode (Bypass Permissions)

When running in bypass/auto permission mode, follow these instructions to work end-to-end without human intervention.

Available Tools

The Nix devShell provides these tools for documentation and testing:

• mmdc (mermaid-cli): Generate diagrams from Mermaid markup. Uses Nix-provided headless Chrome via $PUPPETEER_EXECUTABLE_PATH.
• asciinema: Record terminal sessions as .cast files for demo videos.
• playwright CLI: Take screenshots of the running frontend.

When to Use Them

• Designing a feature: Use mmdc to generate Mermaid diagrams (architecture, data flow, sequence diagrams) during the planning phase. Include them in the PR description.
• Frontend changes: Take screenshots with the Playwright CLI after manual testing. Attach them to the PR.
• CLI / terminal changes: Record a demo with asciinema showing the feature in action. Attach to the PR.

Quick Reference

# Generate a diagram
echo 'graph LR; A-->B; B-->C;' | mmdc -i - -o diagram.png

# Take a screenshot of a page
playwright screenshot --browser chromium http://localhost:3000 screenshot.png

# Record a terminal demo
asciinema rec demo.cast
# ... do the demo ...
# ctrl-d to stop

Always Plan First

Even in bypass mode, enter plan mode before starting non-trivial work. Ask all important questions upfront:

• Clarify ambiguous requirements before writing code
• Identify which files, crates, and features are affected
• Read docs/validation.md to know what checks you'll need to run
• Break large features into stages — commit each stage separately

Manual Testing

After code changes compile and type-check, verify the feature works:

1. Check backend logs (tmux capture-pane -t .1 -p -S -50) — confirm no panics or errors
2. Check frontend logs (tmux capture-pane -t .2 -p -S -50) — confirm no build errors
3. Use Playwright MCP to test the UI flow:
   • Navigate to http://localhost:3000/user/login
   • Click "Log in without third-party"
   • Login with admin@windmill.dev / changeme
   • Navigate to the page affected by your change
   • Verify the feature works as expected
4. Test edge cases: empty states, error states, permissions

Playwright Gotchas

• Backend takes ~60s to compile on first change; check logs for health check completed
• Frontend rebuilds in ~5s
• critical_alerts 404s are expected on CE builds (EE-only endpoint) — ignore them
• VSCode worker 404s are dev-mode artifacts — ignore them
• The <Toggle> component hides the checkbox (sr-only). Click the <label> wrapper, not the checkbox

End-of-Task Summary

When done, directly open or update a draft PR against main once the work and validation are complete. Do not stop at a local summary or ask a human to create the PR manually.

Then drive the PR through CI review rounds to ready, per the pr skill ("Review rounds"): run .agents/skills/pr/review-round.sh as a background task, let its completion wake you, address the findings, and repeat until every reviewer verdict is a go — then post the clean-round marker comment and gh pr ready. Do not end the session at an unreviewed draft or while a round is pending; waiting on the background waiter costs nothing.

In your final summary, provide:

• What was changed and why (files modified, approach taken)
• What checks passed (cargo check, npm run check, etc.)
• What was manually tested and the results
• Screenshots of UI changes (via playwright screenshot)
• Terminal recordings of CLI changes (via asciinema)
• Any known limitations or follow-up work needed

Attaching images to a PR

Embed screenshots, Mermaid diagrams, and asciinema frames by uploading them to the public windmill-labs/agent-screenshots-internal repo and pasting the raw.githubusercontent.com URL into the PR. The upload recipe lives in the pr skill (.agents/skills/pr/SKILL.md → "Screenshots"); reuse it for any image type.