epiphan-house-rules-skill
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines workflows that ingest data from external sources (HubSpot CRM, Epiphan Knowledge, and product catalogs) to generate customer-facing Gmail drafts and internal Slack alerts. This creates a surface where malicious content embedded in the source data could influence the agent's output or actions.
- Ingestion points: Data retrieved from
qualify_lead,search_product_catalog, and Epiphan Knowledgesearchinreference/gates.md. - Boundary markers: No explicit delimiters or instructions are provided to the agent to treat external search results or CRM fields as untrusted data.
- Capability inventory: The skill utilizes
create_draft(Gmail), Slack DMs, and calendar writes as defined inreference/gates.md. - Sanitization: There is no mention of sanitizing or escaping content retrieved from external tools before interpolation into outbound messages.
- [COMMAND_EXECUTION]: The skill requires the agent to perform local file system operations to maintain persistent logs outside of the standard skill environment.
- Evidence:
reference/self-healing.mdinstructs the agent to append JSON log lines to~/.claude/skill-runs/{skill-name}.jsonlfor every execution.
Audit Metadata