fix(openai-agents): Stop destructively wrapping tools

[alexander-alderman-webb]

Description

Overwrite the on_invoke_tool property instead of wrapping the entire tool.

Otherwise, the follow-up PR has the test error

FAILED tests/integrations/openai_agents/test_openai_agents.py::test_tool_definitions[True-True] - agents.exceptions.UserError: ToolSearchTool() requires at least one searchable Responses surface: a tool_namespace(...) function tool, a deferred-loading function tool (`function_tool(..., defer_loading=True)`), or a deferred-loading hosted MCP server (`HostedMCPTool(tool_config={..., 'defer_loading': True})`).

as defer_loading is not forwarded in the wrapper.

Issues

Reminders

• Please add tests to validate your changes, and lint your code using uv run ruff.
• Add GH Issue ID & Linear ID (if applicable)
• PR title should use conventional commit style (feat:, fix:, ref:, meta:)
• For external contributors: CONTRIBUTING.md, Sentry SDK development docs, Discord community

[github-actions]

Codecov Results 📊

✅ 99634 passed | ⏭️ 6757 skipped | Total: 106391 | Pass Rate: 93.65% | Execution Time: 365m 48s

📊 Comparison with Base Branch

Metric	Change
Total Tests	📉 -160
Passed Tests	📉 -135
Failed Tests	📉 -1
Skipped Tests	📉 -24

➖ Removed Tests (1)

View removed tests

• test_continuous_profiler_auto_start_and_stop_sampled_span_streaming[non-experiment-thread]
  • File: tests.profiler.test_continuous_profiler

All tests are passing successfully.

✅ Patch coverage is 100.00%. Project has 2506 uncovered lines.
✅ Project coverage is 89.88%. Comparing base (base) to head (head).

Coverage diff

@@            Coverage Diff             @@
##          main       #PR       +/-##
==========================================
+ Coverage    89.88%    89.88%        —%
==========================================
  Files          193       193         —
  Lines        24686     24753       +67
  Branches      8814      8860       +46
==========================================
+ Hits         22188     22247       +59
- Misses        2498      2506        +8
- Partials      1400      1415       +15

---

Generated by Codecov Action

[sentry]

Bug: The _sentry_wrapped flag is set on the original function, not the wrapper. This causes the idempotency check to fail on subsequent calls, leading to tools being wrapped multiple times.
_{Severity: HIGH}

Suggested Fix

The _sentry_wrapped flag should be set on the new wrapper function after it is created, not on the original function before it is replaced. The correct sequence is to create the wrapper, set the flag on it, and then assign it to tool.on_invoke_tool.

Prompt for AI Agent

Review the code at the location below. A potential bug has been identified by an AI
agent. Verify if this is a real issue. If it is, propose a fix; if not, explain why it's
not valid.

Location: sentry_sdk/integrations/openai_agents/patches/tools.py#L62-L63

Potential issue: The idempotency check for wrapping `FunctionTool` instances is flawed.
The code sets the `_sentry_wrapped` flag on the original `tool.on_invoke_tool` function
before replacing it with a new wrapper function. In subsequent calls within a multi-turn
agent run, the check inspects the new wrapper function, which does not have the
`_sentry_wrapped` attribute. This causes the check to fail, leading to the tool being
wrapped repeatedly. This results in deeply nested spans and incorrect telemetry for any
agent that reuses tools across multiple turns.

_{Did we get this right? 👍 / 👎 to inform future reviews.}

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 27d9ea6. Configure here.}

[cursor]

Stale agent on shared tools

Medium Severity

Permanently replacing on_invoke_tool and skipping already-wrapped tools freezes the agent captured by the invoke wrapper. When the same FunctionTool is reused by another agent (shared tools, clones, or handoffs), later execute_tool spans keep the first agent's name, model settings, and conversation id.

Additional Locations (2)

• sentry_sdk/integrations/openai_agents/patches/tools.py#L30-L33
• sentry_sdk/integrations/openai_agents/patches/tools.py#L55-L56

 

^{Reviewed by Cursor Bugbot for commit 27d9ea6. Configure here.}

[sentrivana]

lgtm, but check out the bot comments if you haven't yet in case there's something to them