product-marketing-context
Pass
Audited by Gen Agent Trust Hub on Sep 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external information to populate the marketing 'brain' file, which introduces a potential attack surface for indirect prompt injection.\n
- Ingestion points: The skill reads from
/foundation/brain.md, legacy files in.agents/and.claude/, and accepts user-pasted documents such as pitch decks and CRM reports.\n - Boundary markers: The skill relies on a structured Q&A format but does not use explicit delimiters or 'ignore' instructions for processed content.\n
- Capability inventory: The skill has permissions to write to the local filesystem (
/foundation/brain.mdand.brain-draft.md).\n - Sanitization: The skill employs 'Operating Rule 1', mandating user confirmation for all file writes, and a 'Quality Gate' that challenges and rejects vague or generic inputs.\n- [DYNAMIC_EXECUTION]: The skill uses a template-based system to generate documents from user-provided variables.\n
- Evidence: The
templates/brain-template.mdfile contains{{variable}}placeholders that are dynamically substituted with user answers at runtime before the final content is written to the workspace.
Audit Metadata