The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes complex BGP topology data (prefixes, session states, AS-paths) from external Topolograph instances. This ingestion of external data creates a surface where maliciously crafted BGP information could attempt to influence the agent's reasoning.
Ingestion points: Data returned by tools like search_bgp_routes, get_bgp_graph, and list_bgp_sessions (SKILL.md).
Boundary markers: The skill includes 'Boundary' and 'Never' sections to define usage constraints, though it lacks specific instructions for sanitizing external data.
Capability inventory: The skill provides read-only tools for network analysis and lacks file-write or subprocess execution capabilities.
Sanitization: No explicit sanitization or filtering of the BGP data is defined.
[DATA_EXFILTRATION]: The skill connects to an external network service at https://topolograph.com/mcp (or a user-defined URL) to perform its functions. This is a vendor-owned resource used for the skill's primary network analysis purpose.
[COMMAND_EXECUTION]: The documentation includes bash code blocks for defenseclaw tool allow commands. While intended for user-driven security configuration, these instructions describe modifications to the agent's tool access permissions.

topolograph-bgp-analysis