The Agent Skills Directory

[PROMPT_INJECTION]: Indirect Prompt Injection Surface: The skill is designed to ingest and analyze untrusted data from various external sources, including business websites, GitHub profiles, and public forum posts (e.g., Reddit, Hacker News). This creates an attack surface where malicious instructions embedded in these public sources could attempt to influence the agent's behavior during the lead qualification and scoring phases.
Ingestion points: External content is fetched through tools like Firecrawl, Browserbase, and browser-assisted research as described in references/data-sources.md and references/demand-signals.md.
Boundary markers: The instructions do not specify the use of delimiters or explicit 'ignore embedded instructions' warnings for external data interpolation.
Capability inventory: The skill has access to network-enabled tools and file-writing capabilities (CSV generation).
Sanitization: No explicit sanitization or escaping of external content is mandated before analysis.
[COMMAND_EXECUTION]: Local Tool Invocation: The skill instructs the agent to execute a local CLI tool (node tools/clis/github-prospects.js) to enrich lead data from GitHub. While this is a functional component, it represents a point of local command execution with user-supplied parameters (repository names).
[EXTERNAL_DOWNLOADS]: Integration with Well-Known Services: The skill references and integrates with several reputable technology and data providers, such as Apollo, ZoomInfo, Clay, and GitHub, for legitimate business operations including data enrichment and email verification. These are recognized as well-known sales technology services.
[SAFE]: Robust Compliance Framework: The skill includes a dedicated compliance reference (references/compliance.md) that strictly forbids risky behaviors such as bulk scraping, bypassing bot protections, and harvesting data from illicit or unprovenanced sources. It provides detailed guidance on adhering to GDPR, CAN-SPAM, and CASL.

prospecting