The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute various shell commands using the membrane CLI tool. These commands include authenticating (membrane login), searching for integrations (membrane search), and executing specific API actions (membrane action run).
[EXTERNAL_DOWNLOADS]: The instructions guide the installation of the @membranehq/cli package from the NPM registry and the use of npx to execute the latest version of the CLI. These resources are provided by the skill's author to interface with their platform.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because it ingests and processes data from external sources (Wonderchat API conversations and user records).
Ingestion points: Untrusted data enters the agent context through the terminal output of membrane action run and membrane request commands.
Boundary markers: The instructions lack specific markers or delimiters to help the agent distinguish between its instructions and potentially malicious content within the retrieved data.
Capability inventory: The skill has the capability to execute further actions and make network requests via the membrane CLI based on processed data.
Sanitization: No sanitization, filtering, or validation steps are defined for the data fetched from the external service before it is presented to the agent.

wonderchat