The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on executing shell commands via the Bash tool to manage Python environments and dependencies.
The agent is instructed to run uv commands such as init, add, sync, and lock to modify the local environment and project structure.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection through the processing of project configuration files.
Ingestion points: The skill processes pyproject.toml, requirements.txt, and uv.lock files, which may be provided by untrusted sources in a project directory (found in SKILL.md and REFERENCE.md).
Capability inventory: The agent uses the Bash tool to perform operations based on these files, including installing packages which could lead to code execution.
Boundary markers: There are no explicit instructions or delimiters provided to the agent to disregard natural language instructions that might be embedded within configuration comments or metadata.
Sanitization: There is no evidence of validation or sanitization of the configuration file content before it is processed by the underlying tools.

uv-project-management