The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests carrier data from a federal census mirror, creating an attack surface for indirect prompt injection. Although this represents a vulnerability surface, it is inherent to the skill's data-fetching purpose.\n
Ingestion points: Carrier information is retrieved via the us-motor-carriers tool (SKILL.md).\n
Boundary markers: The instructions lack explicit delimiters or instructions for the agent to disregard potentially malicious content in the census data.\n
Capability inventory: The skill has access to a bash shell, jq, and the /tmp directory for data processing (SKILL.md).\n
Sanitization: No specific data sanitization or validation steps are defined for the tool output before it is processed by the agent.\n- [COMMAND_EXECUTION]: The skill utilizes a shell environment to execute specialized programs and data processing utilities.\n
Evidence: Instructions provided for using us-motor-carriers, jq, programs list, and programs enable within a bash environment (SKILL.md).\n- [EXTERNAL_DOWNLOADS]: The skill instructions direct users to connect a remote Model Context Protocol (MCP) server for tool access.\n
Evidence: Directs users to https://mcp.hedwigai.com/ and the installation of the @un-seen/hedwigai-skills plugin (SKILL.md).

trucking-fleets