The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill instructions demonstrate passing sensitive authentication data (Authorization_string=$F5_AUTH_STRING) directly as an environment variable prefix in shell commands. This practice can expose credentials to other local users or processes on the system through process monitoring tools (e.g., ps).
[COMMAND_EXECUTION]: The skill relies on shell command execution to interact with the F5 management plane, specifically calling Python scripts defined by environment variables ($MCP_CALL and $F5_MCP_SCRIPT). While standard for this integration, it creates a dependency on the integrity of the environment configuration.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze F5 system logs (show_logs_tool), which are a source of untrusted external data. Attackers can intentionally trigger specific log messages (e.g., via malformed HTTP requests or SSL handshake failures) to influence the agent's reasoning.
Ingestion points: System log output from show_logs_tool in Step 7 and Step 8.
Boundary markers: None; the instructions do not specify delimiters or warnings to ignore instructions embedded within the logs.
Capability inventory: The skill possesses write/delete capabilities through create_tool, update_tool, and delete_tool across all BIG-IP objects.
Sanitization: None; the agent is instructed to scan logs for patterns and correlate events without specific data validation or escaping mechanisms.

f5-health-check