Support Codex pre-tool-use targets

[nikitakarimoffaipro-png]

Summary

Fixes Codex PreToolUse enrichment so Codex-native tool payloads can produce file-scoped memory context instead of falling through the Claude-only tool-name filter.

What changed

• Keeps the existing Claude/Copilot Edit/Read/Grep behavior unchanged.
• Adds conservative Codex adapters for:
  • apply_patch, parsing *** Add/Update/Delete File: patch headers from tool_input.command (the shape observed from Codex Desktop), patch, or input.
  • exec_command / shell_command / Bash, parsing only simple single-command read/search/list forms.
• Rejects compound shell commands containing shell control characters so the hook stays silent when it cannot safely infer a file target.
• Updates hooks.codex.json so Codex-native tool names route to PreToolUse.
• Adds focused tests for Codex patch and shell payloads plus manifest/connect matcher coverage.

Why

The Codex plugin currently registers PreToolUse, but the hook script only recognizes Claude-style tool names (edit, write, read, glob, grep). In a live Codex Desktop session, apply_patch arrived as:

{
  "tool_name": "apply_patch",
  "tool_input": {
    "command": "*** Begin Patch\n*** Update File: ..."
  }
}

That meant the hook exited before /agentmemory/enrich, so file-scoped context injection was effectively inert for Codex patch edits.

Validation

• npm run build
• npx vitest run test/context-injection.test.ts test/codex-plugin.test.ts test/codex-connect-hooks.test.ts
• Live host shadow test against a running agentmemory server:
  • real Codex apply_patch event recorded files: ["/tmp/agentmemory-pretool-live-proof.txt"], wouldInject: true, filteredWouldInject: true, exitCode: 0
  • direct host-wrapper Codex exec_command payload recorded files: ["/tmp/agentmemory-pretool-live-proof.txt"], terms: ["command-shape"], wouldInject: true, filteredWouldInject: true, exitCode: 0

I also ran npm test; it reached 1408 passing tests but failed on pre-existing/environment-sensitive cases unrelated to this change:

• provider/auto-compress defaults in this install environment
• Minimax default-model expectation mismatch
• hook-project tests expecting the checkout basename to be exactly agentmemory while this local clone resolved as agentmemory-upstream

[vercel]

Someone is attempting to deploy a commit to the rohitg00's projects Team on Vercel.

A member of the Team first needs to authorize it.

[coderabbitai]

Important

Review skipped

Draft detected.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: 5cfbe876-0508-4780-8e0b-92a3cfdc264a

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}