proof-points
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function involves processing untrusted source material (sales decks, battlecards, CRM data) which may contain hidden instructions intended to influence the agent.
- Ingestion points: Step 2 (Extract Mode) mines claims from user-provided source collateral and optional connectors (CRM, Chat, Cloud Storage).
- Boundary markers: The skill lacks explicit boundary markers or instructions to treat external data as untrusted text during the extraction process.
- Capability inventory: The skill has permissions to read and append to the project's central knowledge file (
/foundation/brain.md) and session logs (/context/skill-sessions.md). - Sanitization: The skill relies on manual verification flags (
[NEEDS PROOF]) and a final user confirmation prompt in Step 6 rather than technical sanitization or automated filtering of the ingested content.
Audit Metadata