The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to run commands like npx -y @nomadamas/k-skill@0 instruct and npx -y @nomadamas/k-skill@0 update. This involves downloading and executing code from a package on the NPM registry that is not associated with a verified organization.
[INDIRECT_PROMPT_INJECTION]: The skill processes real-time travel data from the MyRealTrip MCP server, creating a potential attack surface where malicious data could influence agent behavior.
Ingestion points: Travel search results (flights, accommodations, tours) fetched via the https://mcp-servers.myrealtrip.com/mcp endpoint in scripts/myrealtrip_mcp.py.
Boundary markers: None identified. The skill does not provide delimiters or instructions to the agent to disregard instructions potentially embedded in the travel data.
Capability inventory: The skill environment supports shell command execution (npx, python3, pip) and network communication via the MCP protocol.
Sanitization: No evidence of sanitization or filtering of the API response data is performed before it is presented to the agent.
[DYNAMIC_EXECUTION]: The script scripts/myrealtrip_mcp.py uses importlib.import_module to dynamically load SDK components. While this is used for version compatibility, dynamic loading from computed strings is a risky pattern.
[DYNAMIC_EXECUTION]: The test suite scripts/test_myrealtrip_mcp.py uses importlib.util and spec.loader.exec_module to load and execute code during testing.
[COMMAND_EXECUTION]: The skill's operational flow relies on the agent executing shell commands to fetch instructions, perform updates, and call the travel search tools.

myrealtrip-search