fix: bind Agent Attention delivery to Codex task

[myagentdojo]

Agent Attention can now record successful Codex delivery using the result that Codex actually returns. The receipt binds threadId to the claimed task, so wrong-task and unbound boolean results fail closed. Existing stored receipts remain readable.

Validation: 70 lifecycle tests passed. The installed payload suite passed 5/5. The clean-candidate harness-install proof passed 46 tests with 3 environment-gated skips and no failures.

Summary by CodeRabbit

• Bug Fixes

  • Delivery confirmations now require proof tied to the exact task thread.
  • Rejected invalid, mismatched, malformed, and legacy delivery claims.
  • Removed acceptance of generic “delivered” confirmations without task-specific proof.

• Tests

  • Added coverage for valid task-specific delivery results.
  • Expanded validation tests for unsupported and unbound proof formats.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 9faac1e5-ba98-4b8d-a1b2-d9bbdfe22a41

📥 Commits

Reviewing files that changed from the base of the PR and between d843f6a and a46595f.

📒 Files selected for processing (3)

• experiments/agent-attention/runtime/agent-attention/agent-attention.py
• experiments/agent-attention/runtime/agent-attention/test_agent_attention.py
• plugin/runtime/agent-attention.py

---

📝 Walkthrough

Walkthrough

record_delivery now accepts delivery only when the tool result is an object with a threadId matching the claimed task thread. Runtime implementations and tests enforce this exact-task proof.

Changes

Exact task delivery proof

Layer / File(s)	Summary
Validate exact task-thread proofs experiments/agent-attention/runtime/agent-attention/agent-attention.py, plugin/runtime/agent-attention.py	Both record_delivery implementations validate the claim before parsing the result. They require a matching threadId and remove the generic delivered: true requirement.
Cover accepted and rejected proofs experiments/agent-attention/runtime/agent-attention/test_agent_attention.py	Tests cover valid matching proofs and reject mismatched, malformed, legacy, alternate-key, and non-object proofs. The expected error text now requires exact task delivery confirmation.

Estimated code review effort: 2 (Simple) | ~10 minutes

Possibly related PRs

• myagentdojo/agent-plugin-playground#1: Introduces the record_delivery implementation and tests refined by this change.
• myagentdojo/agent-plugin-playground#3: Also updates record_delivery and its tests to enforce threadId task binding.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and concisely describes the main change: binding Agent Attention delivery to the claimed Codex task.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch codex/fix-agent-attention-delivery-receipt

---

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