The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines workflows that ingest large amounts of untrusted data from the open web, creating a surface for indirect prompt injection.
Ingestion points: The skill uses tools like last30days, social-fetch, Firecrawl, and Browserbase to fetch content from Reddit, Hacker News, GitHub issues, and general business websites (referenced in references/demand-signals.md and references/data-sources.md).
Boundary markers: There are no instructions provided to the agent to use delimiters or ignore embedded instructions when processing data from these external sources.
Capability inventory: The skill allows the agent to perform network scraping operations, generate structured lead sheets (CSV/Markdown), and draft outreach content for subsequent stages.
Sanitization: The instructions do not specify any sanitization or validation of the retrieved text before it is used to qualify candidates or generate outreach rationale.
[COMMAND_EXECUTION]: The skill references the use of a local CLI tool (node tools/clis/github-prospects.js) to process GitHub data. While this is a local utility, it involves executing shell commands with arguments (repository names) derived from user or data-driven input.

prospecting