The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Recommends installing the @membranehq/cli package via npm, which is the official tool provided by the vendor for platform interactions.
[COMMAND_EXECUTION]: Uses the membrane CLI to manage connections, search for actions, and execute integration logic. These operations are within the scope of the skill's intended purpose.
[CREDENTIALS_UNSAFE]: Includes positive security instructions that explicitly tell the agent not to request API keys or tokens from the user, relying instead on the platform's managed authentication flow.
[PROMPT_INJECTION]: The skill processes data from Semaphore actions, creating a potential surface for indirect prompt injection.
Ingestion points: External data from Semaphore enters the context through the output of membrane action run and membrane action list in SKILL.md.
Boundary markers: No specific delimiters or warnings for the agent to ignore instructions within the data payload are present.
Capability inventory: The skill can execute platform-managed actions (membrane action run) and create new ones (membrane action create).
Sanitization: There are no instructions for sanitizing or validating the API output before it is parsed by the agent.

semaphore