The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external data from sources like Apollo job postings and web search results to generate personalized email sequences and call scripts. This content is interpolated into outbound communication templates, creating a surface for indirect prompt injection if external sources contain malicious instructions.
Ingestion points: Job postings via apollo_organizations_job_postings, peer customer descriptions via search_companies_by_description, and general web search gaps.
Boundary markers: The skill requires an explicit AskUserQuestion (Stage 3) for human approval before any outreach is sent or contacts are enrolled.
Capability inventory: apollo_contacts_create, apollo_emailer_campaigns_add_contact_ids, and gmail_create_draft.
Sanitization: Implements check_my_copy (Gate B) and search_product_catalog (Gate S) to verify claims and brand consistency before presentation.
[DATA_EXFILTRATION]: The skill reads from local environment paths (e.g., ~/.claude/skills/epiphan-house-rules-skill/reference/gates.md) to retrieve identifiers and house rules. While it pushes data to external services (Apollo and Gmail), these are the primary functional targets for the sales outreach use case and operate within a human-approved workflow.

prospect-research-to-cadence-skill