The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources such as forums, product reviews, and GitHub issues, which creates a surface for indirect prompt injection attacks.
Ingestion points: Public discourse threads, community posts, and product reviews mined in the demand-signal branch to identify prospect pain points (referenced in references/demand-signals.md).
Boundary markers: The skill contains explicit safety instructions directing the agent to treat fetched content as untrusted data and to never follow instructions embedded within that content (references/demand-signals.md).
Capability inventory: The agent possesses the capability to write lead sheets to the file system (CSV/Markdown) and perform network operations via external tool integrations (SKILL.md, references/data-sources.md).
Sanitization: Instructions recommend paraphrasing over quoting and specifically require escaping all values when generating HTML-based evidence reports (references/demand-signals.md).
[COMMAND_EXECUTION]: The skill documents the use of a vendor-provided CLI tool, github-prospects.js, intended for extracting and enriching data from GitHub stargazers and forks (references/saas-prospecting.md, references/data-sources.md).
[EXTERNAL_DOWNLOADS]: The skill integrates with and references numerous third-party B2B data platforms and search services, such as Apollo, Clay, ZoomInfo, and Truelist, typically via MCP servers or official APIs (SKILL.md, references/data-sources.md). These are identified as legitimate tools for prospecting and enrichment purposes.

prospecting