The Agent Skills Directory

[PROMPT_INJECTION]: The skill ingests untrusted data from the Matrix network. \n * Ingestion points: The scripts matrix-read.py and matrix-read-e2ee.py retrieve message content from external Matrix rooms. \n * Boundary markers: No specific delimiters or protective instructions are implemented for the retrieved message content. \n * Capability inventory: The skill can send messages, edit/redact messages, and perform authenticated API calls to homeservers. \n * Sanitization: No sanitization is performed on incoming message content before it is processed by the AI agent. \n- [COMMAND_EXECUTION]: The diagnostic script performs package installation. \n * The matrix-doctor.py script executes pip install via subprocess.run to manage the matrix-nio library when run with the --install flag. \n- [REMOTE_CODE_EXECUTION]: The codebase includes a reference to a remote execution pattern within a help string. \n * Evidence: scripts/matrix-doctor.py contains a print statement recommending curl -LsSf https://astral.sh/uv/install.sh | sh. This is a recommendation for manual user installation and is not automatically executed by the skill code. \n- [EXTERNAL_DOWNLOADS]: The skill downloads dependencies and interacts with remote homeservers. \n * The skill fetches the matrix-nio[e2e] package from the standard Python Package Index (PyPI). \n * The skill makes authenticated network requests to remote homeservers via the Matrix Client-Server API.

matrix-communication