Add Trustabl security scanning to CI

[joshua-trustabl]

We came across your repo and we like how you leverage real world production traces to optimize agent performance. We scanned the repo, and noticed agent runtime reliability findings that might be worth reviewing.

1. [HIGH] Agent wires MCP servers without input_guardrails
   File: demo/appworld/experiments/code/openai_agents/run.py
   What it means: The agent imports tools from one or more MCP servers (mcp_servers=) but has no input_guardrails.

2. [HIGH] Tool body spawns a subprocess
   File: demo/openai-agents-sdk-demo/agent.py
   What it means: The @function_tool body spawns an OS process.

3. [HIGH] Privileged tool has no needs_approval gate
   File: demo/openai-agents-sdk-demo/agent.py
   What it means: This @function_tool shells out, executes dynamic code, or writes the filesystem, but sets no needs_approval kwarg.

Recommendations are based on our understanding of agent runtime reliability, some findings may be intentional. Please let us know if this was intentional or if our findings are helpful so we can improve the accuracy of the scanner.

Best,
Trustabl.ai
Open-source AI agent reliability scanner (runs locally, GitHub Action)

---

Note

Low Risk
Only adds a third-party GitHub Action workflow with elevated PR/security permissions; no application or runtime code changes.

Overview
Adds a new GitHub Actions workflow .github/workflows/trustabl.yml that runs the Trustabl agent reliability scanner on every push to main and on pull requests.

The job checks out the repo and invokes trustabl/trustabl-action@v0. Workflow permissions grant security-events: write and pull-requests: write (along with read-only contents) so the action can report findings to GitHub security/PR surfaces.

^{Reviewed by Cursor Bugbot for commit 0dde71c. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

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

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

Want higher recall? High effort reviews run extra passes and find more bugs. A team admin can switch effort levels in the Cursor dashboard.

^{Reviewed by Cursor Bugbot for commit 0dde71c. Configure here.}

[cursor]

Scan fails CI by default

High Severity

trustabl/trustabl-action@v0 is invoked with no inputs, so it uses the default severity-threshold of medium and fails the job on medium-or-higher findings. This repository already has HIGH findings in demo agent code (including patterns the PR notes may be intentional), so every push to main and every pull_request will fail this check.

 

^{Reviewed by Cursor Bugbot for commit 0dde71c. Configure here.}

[francescov1]

@joshua-trustabl please dont shill your own product with AI slop PRs