The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection, particularly in the 'Demand-signal' branch which involves ingesting untrusted data from public forums, reviews, social media threads, and GitHub issues.
Ingestion points: The skill explicitly instructs the agent to mine content from external public discourse (e.g., Reddit, Hacker News, X, GitHub) to identify potential customers.
Boundary markers: The instructions provide mitigation strategies by advising the agent to 'quote minimally, paraphrase by default,' and to clearly distinguish between 'observed evidence vs. inference' in reports.
Capability inventory: The agent has the ability to write files (CSV lead sheets) and utilize network tools via MCP integrations (Firecrawl, Browserbase, Apollo) to process this data.
Sanitization: The skill recommends using a JSON-to-HTML generator for reporting that includes escaping all values to prevent cross-site scripting or other injection issues in the generated output.
[EXTERNAL_DOWNLOADS]: The skill references several external B2B data tools and web extraction services (Apollo, ZoomInfo, Clay, Truelist, Firecrawl, Browserbase) via Model Context Protocol (MCP) integrations. These are well-known services within the sales technology ecosystem and are used for their intended primary purposes. The skill also mentions a 'bundled' script, github-prospects.js, for developer-intent prospecting, which is treated as a vendor-provided utility.

prospecting