The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted startup information, including founder materials, traction facts, and cap tables as described in Step 1 of SKILL.md, to generate strategic outputs like narrative memos and round specifications. It lacks explicit instruction boundary markers or sanitization for this ingested data. The skill possesses the capability to search a local knowledge base using the kb search tool as documented in the Corpus deep-dive section of SKILL.md. The combination of ingesting external data and using that information to generate complex strategic advice creates a surface for indirect prompt injection, where malicious instructions within user-provided documents could potentially influence the agent's logic.

vc-strategy