Skip to content

post-commit-reviews.md

Latest commit

 

History

History
81 lines (59 loc) · 3.17 KB

post-commit-reviews.md

File metadata and controls

81 lines (59 loc) · 3.17 KB
Raw

Automation: hands-off reviews

roborev is built to run hands-off. There are two automation layers - turn on both for the full loop.

How roborev works{ loading=lazy }

Layer 1 - Post-commit reviews

A git post-commit hook reviews every commit in the background. This works with any editor or agent.

roborev init      # installs the hook, starts the daemon, registers the repo

The daemon coalesces duplicate hook requests for the same repository, resolved Git reference, and review target. Concurrent editor, worktree, or agent hooks therefore launch one automatic review. Explicit roborev review commands still start a fresh review when you deliberately request one.

Now every commit you make is reviewed automatically. Each review gets a verdict (pass or fail) and, when it fails, a list of findings with severities and file locations. Check that it is running:

roborev status        # daemon + queue
roborev show HEAD     # print the latest commit's review in the terminal

Then act on the reviews in whichever way fits how you work:

  • Copy-paste from the TUI. roborev tui shows the review queue; open a review to read its findings and copy the full text straight into your coding agent (or fix by hand). This works with any agent or editor.
  • Fix failing reviews with the roborev-fix skill. From inside Claude Code or Codex, /roborev-fix (Codex: $roborev-fix) pulls every open failing review for your current branch or git worktree, applies the fixes, and closes the reviews in one pass.
  • Clean the whole branch before a PR with the refine loop. /roborev-refine (Codex: $roborev-refine) reviews the branch, fixes findings, and re-reviews until every review passes.

The roborev-fix and roborev-refine skills come from roborev skills install (see Agent Skills). Agent Hook installation updates the bundled skills automatically for supported profiles.

Layer 2 - Agent hook

The agent hook watches supported coding-agent sessions and, once review work piles up, supplies exact review IDs to the roborev-fix skill before the session ends. It never runs the separate roborev fix --open agent workflow.

roborev agent-hook install    # wire harnesses and update supported bundled skills

See Agent Hook for thresholds and configuration.

Why CLI, not Desktop?

The agent hook relies on harness lifecycle hooks supplied by Claude Code, Codex, Copilot CLI, Cursor, Factory Droid, Gemini CLI, Hermes, and Qwen. Claude Desktop does not expose these hooks, so Layer 2 does not run there. Layer 1 (post-commit reviews) works regardless of which agent or app you use.

Let an agent finish setup

Point your coding agent at the built-in guide and it will inspect this repo and help you finish configuration:

roborev quickstart            # human-readable
roborev quickstart --json     # machine-readable state for agents

React to review events

To notify a channel or file issues automatically when reviews complete, add Review Event Hooks (desktop notifications, Slack, Kata, webhooks, or any shell command).