Hackathon Loop
Run the following steps in an infinite loop until the user says "stop":
Claude CodeGeneric
Run the following steps in an infinite loop until the user says "stop":
**Before each iteration:** Read ./loop-logs/summary-*.md to load memory from previous iterations. Note the iteration number and increment it.
1. **PRD** — Read ./prd.md if it exists. Write an improved PRD to ./prd.md targeting:
- https://global-fusion-hackathon.devpost.com/
- https://orion-build-challenge.devpost.com/
- https://techbuilders.devpost.com/
Use sequential-thinking MCP for complex architectural decisions.
Use context7 MCP to verify your tech stack choices are current.
2. **Verify** — Review the PRD as a strict hackathon judge. Check every item in the PRD quality bar from CLAUDE.md. Rewrite immediately if it fails.
3. **Implement** — Execute the PRD. Spawn parallel subagents for frontend and backend if the work is large. Commit: `git add -A && git commit -m "feat: iteration N — summary"`
4. **Test** — Run all existing tests. Write new Playwright e2e tests for every user flow. Run `npx tsc --noEmit` and `npm run lint`. Fix ALL failures before continuing. Log results to ./loop-logs/test-results-N.txt
5. **Status + Design** — Write status to ./loop-logs/status-N.md. Use Figma MCP to create/update all screens. Save Figma URL to ./figma-url.txt. Apply design improvements to code.
6. **Compact** — Write one-paragraph summary to ./loop-logs/summary-N.md. Run /compact. Then immediately restart at Step 1.
Never stop unless the user says "stop" or "exit".
Maintain Hackathon Loop?
Let people know it's listed here — add the badge (live metrics, light/dark aware) or a plain link to your README or docs.
[Hackathon Loop on getagentictools](https://getagentictools.com/loops/aaravjj2-hackathon-loop?ref=badge) npx agentictools info loops/aaravjj2-hackathon-loop The second line is the CLI lookup for this page — handy in READMEs and docs.