The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external IndyKite MCP endpoints into the agent's context, creating a potential surface for indirect prompt injection if the source data is manipulated.
Ingestion points: The agent receives data from the IndyKite MCP server via tools/call (AuthZEN decisions) and resources/read (Knowledge Query results) methods, which are handled by the scripts/mcp-call.sh script.
Boundary markers: The skill does not provide specific instructions to the agent on how to isolate or delimit external API responses from the rest of the prompt context.
Capability inventory: The skill utilizes bash and curl to interact with external network services and read local environment variables (BEARER_TOKEN, PROJECT_GID).
Sanitization: There is no explicit requirement or method documented for sanitizing or validating the JSON-RPC results before they are incorporated into the agent's reasoning process.
[COMMAND_EXECUTION]: The skill includes a bash script (scripts/mcp-call.sh) that executes curl to perform network operations. The script implements basic regex validation for command-line arguments (method and name) to prevent command injection from user-supplied inputs.

indykite-mcp-server