# Project self-check

AI Workshop · episode 10: Viewer project review
https://mihalkevich.com/en/workshop/viewer-review

- [ ] A newcomer (or an agent) runs the project from CLAUDE.md without questions
- [ ] The agent knows how to verify its work, and does
- [ ] Secrets aren't in the repo, the agent's permissions are minimal
- [ ] Tasks follow the template: goal, context, constraints, definition of done
- [ ] There's at least one automated harness layer: a hook, a test or an eval set
- [ ] You understand every line you ship to production

## Harness in this episode

The final review tests the harness on someone else's project: which layers exist and which to add.

## Official documentation

- [Claude Code — best practices](https://code.claude.com/docs/en/best-practices)
- [Anthropic — Effective harnesses for long-running agents](https://www.anthropic.com/engineering/effective-harnesses-for-long-running-agents)
- [Anthropic — Demystifying evals for AI agents](https://www.anthropic.com/engineering/demystifying-evals-for-ai-agents)

— mihalkevich school