The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill installs the @membranehq/cli package from the official NPM registry. This package belongs to the skill's authoring organization and is used for its intended purpose of managing cloud integrations.\n- [COMMAND_EXECUTION]: The skill executes various membrane CLI commands to authenticate users and interact with the Medius service. These operations are restricted to the vendor's own command-line interface.\n- [METADATA]: A documentation link for 'Oculus' is included in the 'Medius' skill. This appears to be a copy-paste error from another skill template and does not point to a malicious domain.\n- [INDIRECT_PROMPT_INJECTION]: The skill interpolates user-provided queries into shell commands (e.g., the --intent and --description flags). While this presents a surface for indirect prompt injection, it is mitigated by the fact that these commands are executed within the context of the user's authenticated Membrane session.\n
Ingestion points: User-provided 'QUERY' and 'DESCRIPTION' strings in SKILL.md shell examples.\n
Boundary markers: None present in the command examples.\n
Capability inventory: Subprocess execution via membrane action list and membrane action create.\n
Sanitization: No explicit sanitization or escaping logic is described in the prompt instructions.

medius