python-sdk
Warn
Audited by Gen Agent Trust Hub on Mar 4, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: Multiple documentation files, including 'references/tool-builder.md' and 'references/agent-patterns.md', provide code examples that use the
eval()function to process tool arguments (e.g.,eval(call.args['expression'])). If followed, these patterns allow for arbitrary code execution on the host environment using input provided by the AI agent or user. - [COMMAND_EXECUTION]: The skill defines an allowed tool permission for
Bash(python *), which permits the execution of arbitrary Python scripts. While intended for SDK usage, this broad permission allows for the execution of dynamically created or fetched code. - [EXTERNAL_DOWNLOADS]: The skill instructions initiate the installation of the 'inferencesh' package from an external registry. It also demonstrates patterns for fetching remote resources from external domains such as 'example.com' and 'hooks.slack.com'.
Audit Metadata