agent-ready-codebaselisted

# Agent-Ready Codebase **Source**: AI Builder Club, "Loop Engineering Guide (2026)," which frames this as the most frequently overlooked of the "four ingredients for compounding loops" (attributed to AI Jason's "Loop Engineer: Systemization and Artifacts" framework), despite determining whether autonomous agent operation is feasible at all. ## The three properties **Legible — agents can locate what needs changing.** - Maintain a compact (~100-line) index document (`AGENTS.md` or `CLAUDE.md`) that orients an agent quickly, rather than requiring it to explore the whole tree every session. - Implement custom lints that warn against discouraged patterns specific to the codebase — encoding tribal knowledge as a checkable rule instead of leaving it undocumented. **Executable — work begins at near-zero token cost.** - Pre-boot development servers in setup scripts, so an agent doesn't spend its first several turns just getting the environment running. - Structure the repo to be worktree-friendly, so multiple agents can work in parallel without stepping on each other's uncommitted state. - Provide scenario-jumping scripts for rapid state transitions — a way to get straight to the relevant application state instead of manually navigating there every time. **Verifiable — tools exist to prove outcomes, not just claim them.** - Browser automation with recording (e.g., Playwright with video capture) so a UI-affecting change has actual evidence, not a self-report. - End-to-end tests th