The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructions direct users to install its FortiOS analysis capabilities from a personal GitHub fork (github.com/akshaysiddaram/fwrule-mcp.git) instead of the main project repository or a verified package registry. Installing and executing code from unverified personal accounts poses a significant supply chain risk, as the content of the fork is not subject to the same oversight as the primary repository.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process large, complex firewall configurations (ruleset_payload, candidate_rule_payload) which serve as an untrusted data source. This content could be crafted to contain hidden instructions intended to influence the agent's behavior during the analysis process.
Ingestion points: The ruleset_payload, candidate_rule_payload, and context_objects parameters in the analyze_firewall_rule_overlap and parse_policy tools.
Boundary markers: Absent. The skill instructions do not specify any delimiters or safety prompts to isolate the firewall configuration data from the agent's instructions.
Capability inventory: The skill requires python3 and relies on an external MCP server (fwrule-mcp). While the skill is described as "read-only," the execution environment has access to the agent's standard toolset.
Sanitization: No sanitization, validation, or instruction-filtering is mentioned for the raw vendor-native configuration text before it is processed by the underlying analysis engine.

fwrule-analyzer