[ajianaz]

What

Part 3 of 3 of the child-agent handoff: the real-binary integration test from the issue's testing section.

Why

#383. The wiring merged in #397 is covered by unit tests; this proves the whole path with real child processes.

Changes

• tests/child_handoff_e2e.rs (test only): the test plays the parent model with AgentStartTool / AgentPollTool and starts real tole children through a wrapper that pins a scripted mock provider and a no-op uteke (hermetic: the real memory store is never touched).
  1. an implement child submits a handoff whose verification failed: agent_poll reports verify_failed (not closed) with a fix prompt naming the command and the artifact;
  2. a fix child started from that prompt reads the artifact by reference with its own file tool (the mock only lets it report saw artifact if it received the file body), submits passing verification: closed, no fix prompt;
  3. an implement child that closes without any handoff is unverified.

Testing

• The new test passes locally (about 4 s); fmt, clippy -D warnings clean; cora review --staged clean.
• Not done: a live run with real agents on a real provider; feat(child-agents): typed handoff + mandatory verify gate for child agents #383 stays open for that.

Refs #383

[github-actions]

🔍 Cora AI Code Review

✅ No issues found. Code looks good!

---

Review powered by cora-code · BYOK · MIT