The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The scripts/search_plugins.py script retrieves marketplace metadata and repository information from GitHub's API. These network operations target a well-known service and are essential for the skill's functionality to provide up-to-date search results.
[COMMAND_EXECUTION]: Subprocess calls to curl are used for data retrieval. These calls use list-based arguments with subprocess.run, which is a secure pattern that prevents shell injection vulnerabilities.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface because it ingests and displays third-party plugin descriptions. This data enters the agent's context and could contain instructions designed to mislead the AI.
Ingestion points: Third-party metadata is fetched from various repositories listed in scripts/marketplaces.json and processed by the search script.
Boundary markers: No explicit delimiters or instructions are used to distinguish third-party metadata from authoritative skill instructions.
Capability inventory: The skill provides search and metadata retrieval, while the host agent environment likely possesses broader capabilities such as file system access.
Sanitization: The script performs JSON validation but does not sanitize or filter the natural language content within the plugin descriptions.

plugin-search