The Agent Skills Directory

[COMMAND_EXECUTION]: The SKILL.md workflow instructs the agent to execute a shell command: python3 scripts/add_torrent.py "<magnet_uri>". Because the <magnet_uri> is a value retrieved from an external search tool, this pattern creates a risk of command injection if the input contains shell metacharacters or closing quotes that could allow for execution of arbitrary commands.
[REMOTE_CODE_EXECUTION]: The mcp_config.json configuration uses uvx to install and run the torrent-search-mcp package at runtime. Downloading and executing code from public registries without version pinning or source verification poses a supply chain risk.
[DYNAMIC_EXECUTION]: The skill relies on string interpolation to assemble and execute shell commands at runtime. The agent is directed to take output from one tool and pass it directly as a command-line argument to a local script without explicit sanitization steps defined in the instructions.
[INDIRECT_PROMPT_INJECTION]: The skill processes search results from an external service which could contain malicious metadata designed to influence the agent's behavior.
Ingestion points: Data enters via the search_torrents tool output.
Boundary markers: None identified in the instructions or script.
Capability inventory: The skill can execute local Python scripts via the shell and make network POST requests to local IP addresses.
Sanitization: There is no logic provided to sanitize or validate the magnet URIs before they are passed to the shell or the requests library.

torrent-movie