The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes shell command blocks to execute Python scripts for network device interaction. It relies on environment variables such as $MCP_CALL, $PYATS_MCP_SCRIPT, and $PYATS_PYTHON to determine the execution path and interpreter. This is standard behavior for tools integrating with the Model Context Protocol (MCP).
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze output from network routing protocols (OSPF, BGP, EIGRP), which constitutes external, untrusted data.
Ingestion points: Data enters the context via the output of pyats_run_show_command across all sections of SKILL.md.
Boundary markers: No explicit delimitation or 'ignore instructions' warnings are provided for the device output.
Capability inventory: The skill has the capability to execute shell commands using python3 via the $MCP_CALL variable.
Sanitization: There is no evidence of sanitization or filtering of the device output before the agent processes it.
[DATA_EXPOSURE]: The skill references the PYATS_TESTBED_PATH environment variable. In the context of pyATS, this file typically contains sensitive information, including network device IP addresses, credentials, and connection parameters. Accessing this file is a functional requirement for the skill and no evidence of exfiltration was found.

pyats-routing