The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill implements a 'Read-only Python aggregation' feature as described in code-execution.md. This allows the agent to execute Python code via a host kernel using the moviepilot_tools library. While documented as read-only, runtime code generation and execution constitute a significant dynamic execution vector.
[EXTERNAL_DOWNLOADS]: The plugin.md contract includes operations such as plugin.install and plugin.source.install which download and install software from external repositories provided via the repo_url parameter. This allows for the introduction of external code into the host environment.
[COMMAND_EXECUTION]: The skill facilitates high-impact system changes, including process restarts (system.restart) and system-wide upgrades to development branches (system.upgrade.dev), which involve the execution of critical system-level commands.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a substantial surface area for indirect prompt injection attacks.
Ingestion points: Untrusted data enters the agent's context through torrent search results (search.md), site RSS feeds and resources (site.md), and media metadata scraping (media.md).
Boundary markers: No specific delimiters or boundary markers are instructed for use when the agent handles or interpolates these external strings.
Capability inventory: The skill provides access to destructive and high-privilege tools, including file deletion (storage.md), plugin lifecycle management (plugin.md), and system configuration updates (config.md).
Sanitization: There are no requirements for sanitizing, escaping, or filtering external content before it is processed or used in subsequent operations.

moviepilot-api