The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill involves gathering information from untrusted external sources such as LinkedIn, company websites, and press news (Phases 2 and 3). * Ingestion points: Web search tool calls during Phase 2 (Contact-Level Research) and Phase 3 (Account-Level Research). * Boundary markers: The instructions do not include explicit boundaries or warnings for the agent to disregard instructions potentially embedded in external web content. * Capability inventory: The agent's capabilities are restricted to text synthesis and brief generation; no tools for filesystem modification, shell access, or arbitrary network exfiltration are present. * Sanitization: No sanitization of ingested external data is mentioned. The impact of any potential injection is low due to the lack of privileged capabilities.
[NO_CODE]: The skill does not include any executable code, scripts, or package dependencies. It consists entirely of natural language instructions for research and reporting.

prospect-research