The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill installs the @membranehq/cli package from the official NPM registry to facilitate communication with the integration platform.
[DATA_EXPOSURE]: The skill follows secure patterns by specifically instructing the agent not to handle raw credentials (tokens or API keys). Instead, it uses a centralized connection manager to handle authentication server-side.
[COMMAND_EXECUTION]: The skill uses a command-line interface to manage connections and execute actions. These commands are scoped to the functionality of the PingBell integration.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface where it ingests untrusted user data to search for or dynamically create new actions.
Ingestion points: User-provided natural language strings passed to the --intent and DESCRIPTION parameters in SKILL.md.
Boundary markers: No specific delimiters are used to wrap the user-provided queries.
Capability inventory: The skill can execute subprocesses via the membrane action run command across various scripts.
Sanitization: No explicit sanitization or validation of the input strings is performed within the skill instructions.

pingbell