The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs shell command execution to interact with the local environment and the Membrane platform.
Evidence: Frequent use of npx @membranehq/cli and membrane for logging in, searching, and running actions.
[CREDENTIALS_UNSAFE]: The skill manages sensitive authentication tokens by storing them in a predictable local directory.
Evidence: Authentication credentials are saved to ~/.membrane/credentials.json for persistence across sessions.
[EXTERNAL_DOWNLOADS]: Utilizes external packages from the NPM registry to provide its core functionality.
Evidence: Downloads and executes the @membranehq/cli package via npx.
[DATA_EXFILTRATION]: Transmits operational data and authentication information to the vendor's cloud infrastructure.
Evidence: Communication is directed to https://api.getmembrane.com/*, including action inputs, connector parameters, and agent prompts.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a broad vulnerability surface for indirect prompt injection due to its integration with a vast number of external services.
Ingestion points: Data entering the agent context via API responses from 100,000+ external apps (e.g., Slack, HubSpot) through action run and agent-session commands.
Boundary markers: The skill instructions do not specify the use of delimiters or 'ignore' instructions when processing content from external API outputs.
Capability inventory: The skill has extensive capabilities including shell command execution, file system access for credentials, and network operations.
Sanitization: No explicit logic is defined to sanitize or validate the content returned from external APIs before it is processed by the agent.

integrate-anything