The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze potentially untrusted data provided by users, such as startup visions, assumptions, and interview results, which could serve as a vector for hidden instructions.
Ingestion points: User-provided venture descriptions, leap-of-faith assumptions, and customer research data processed via instructions in SKILL.md.
Boundary markers: Present; the skill defines a rigid Output Template and a Verification checklist to constrain agent behavior and response structure.
Capability inventory: The execution environment is restricted to Read, Grep, WebSearch, and WebFetch tools as specified in the SKILL.md frontmatter.
Sanitization: Present; the skill includes explicit warnings to avoid processing personal identifiers, private financials, or confidential roadmap details unless explicitly permitted.
[EXTERNAL_DOWNLOADS]: The skill references external methodology resources from theleanstartup.com and leanstartup.co to ground the agent's logic in industry-standard principles.

lean-startup