devtu-code-optimization

Pass

Audited by Gen Agent Trust Hub on Mar 8, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill requires the execution of several shell commands to validate code quality and functionality. These include ruff check for linting, python -c for verifying imports, and a project-specific tooluniverse.cli for executing tool logic with provided arguments.
  • [EXTERNAL_DOWNLOADS]: The workflow incorporates git fetch and gh pr view commands to interact with the remote repository on GitHub. These operations are used to synchronize state and verify pull request status within the mims-harvard/ToolUniverse repository.
  • [PROMPT_INJECTION]: The skill includes a testing phase that processes real_args_json via a CLI tool, which introduces a surface for indirect prompt injection. If the data passed to the CLI is sourced from untrusted external inputs, it could potentially influence the agent's behavior during the execution of the tool tests.
  • Ingestion points: The <real_args_json> input field used in the tooluniverse.cli run command.
  • Boundary markers: No delimiters or specific safety instructions for argument interpolation are mentioned.
  • Capability inventory: The agent has the capability to execute subprocesses (Python, Git, GitHub CLI) and perform file system operations.
  • Sanitization: The provided instructions do not specify any validation or sanitization routines for the data used in CLI tests.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 8, 2026, 04:54 PM