The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from external prospect websites, which is a vector for indirect prompt injection.\n
Ingestion points: The references/research.md playbook instructs the agent to crawl 3–5 key pages of prospect sites (About, Services, Contact, Blog, etc.) to extract firmographics and tech context.\n
Boundary markers: The agent is required to establish an Ideal Customer Profile (ICP) before research begins and must mark all unverified fields as assumptions.\n
Capability inventory: According to the frontmatter and agents/openai.yaml, the agent has access to bash, run_command, write_to_file, replace_file_content, and view_file.\n
Sanitization: The skill mandates the use of confidence levels (HIGH, MEDIUM, LOW) for all data and requires an 'Unverified' section for inferred or third-party data to prevent the agent from treating untrusted external content as definitive fact.

gtm