The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process data from various external sources, including SEC Form D filings, USPTO trademark applications, and SSL certificate logs. These public records could potentially be manipulated by third parties to include hidden instructions aimed at influencing the AI agent's behavior.
Ingestion points: The skill uses commands like webcertificates search, us-trademarks search, and funding search to fetch external data into the agent's context in SKILL.md.
Boundary markers: There are no specific instructions for the agent to use delimiters or ignore instructions embedded within the retrieved data.
Capability inventory: The agent has access to a bash tool (described as a read-only shell) and a create_workbook tool for data aggregation.
Sanitization: The instructions do not include mechanisms for filtering or sanitizing the content retrieved from public registers.
[COMMAND_EXECUTION]: The skill operates by executing specialized programs and utilities within a shell environment.
Evidence: The instructions in SKILL.md direct the agent to use a bash shell for running tools like traffic, rank, and funding, and suggest using jq for JSON processing and /tmp for temporary storage.
[EXTERNAL_DOWNLOADS]: The skill provides instructions for connecting to an external MCP server and installing necessary plugins.
Evidence: SKILL.md references the vendor's domain https://mcp.hedwigai.com/ and marketplace installation commands like /plugin install hedwigai@hedwigai. These are documented as part of the intended setup process for the vendor's tools.

company-signals