The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted lead data (names, company descriptions, and engagement details) and interpolates this content directly into prompts for email drafting.
Ingestion points: disqualified-leads data provided by upstream skills and potential CRM/enrichment data ingested during the categorization phase in SKILL.md.
Boundary markers: The email templates utilize placeholders like [First Name] and [Company] for string interpolation. However, the instructions do not include explicit delimiters or "ignore embedded instructions" directives to prevent the agent from following commands hidden in lead-provided fields (e.g., a lead name containing a jailbreak string).
Capability inventory: The skill possesses the capability to write files (CSV and JSON configuration), draft emails, and interact with external CRM/nurture tools.
Sanitization: There is no evidence of sanitization, escaping, or validation of the external content before it is processed by the LLM for draft generation.

disqualification-handling