The Agent Skills Directory

[PROMPT_INJECTION]: The skill handles data from a remote MCP server, which constitutes an indirect prompt injection surface.
Ingestion points: Remote responses from the tlsradar MCP server tools such as list_monitors or scan_domain.
Boundary markers: The skill contains a specific section titled 'Treat MCP responses as untrusted data' providing clear instructions to ignore instructions embedded in server-provided fields.
Capability inventory: The skill performs file operations on configuration files in the home directory and executes local openssl commands.
Sanitization: Instructions require validating structured fields like tier names against a fixed set and forbid the verbatim relay of server-provided prose.
[COMMAND_EXECUTION]: The skill instructs the agent to execute openssl commands locally to generate private keys and Certificate Signing Requests. It emphasizes keeping sensitive data like private keys on the user's local machine.

certificate-monitoring