Crit — GitHub-style Inline Comments for Agents

Main

• Use Crit after each agent iteration: leave line-level comments, click Finish, and let the agent respond with a targeted patch instead of re-reading entire files.

• Prefer the /crit command integration when your agent supports it; it automates the wait-review-act loop.

• Keep agent_cmd configured only in ~/.crit.config.json (README security note) to avoid repo-level command injection.

Source-backed notes

• README shows install options including brew install crit and go install ...@latest.
• README lists core commands: crit, crit plan.md, crit status, crit cleanup, plus PR helpers like crit push/crit pull.
• README includes a security note: project-level configs cannot set agent_cmd; only the global config can, to prevent malicious repos executing commands.

FAQ

• Does it work only with Claude?: No — README says it works with many agents; it’s a review UI + loop.
• Can I review git diffs automatically?: Yes — crit without args auto-detects changed files and shows diffs.
• Is it safe to enable live agent execution?: Use it only on trusted repos; follow the README security model for agent_cmd.