marketing-os
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data sources, creating a potential surface for indirect prompt injection attacks where malicious instructions could be embedded in ingested content.\n
- Ingestion points: Consumes data from customer research, reviews, call transcripts, and public market sources as described in references/evidence-and-context.md.\n
- Boundary markers: Utilizes a 'truth file' and an evidence-labeling system (Verified, User-provided, Hypothesis) to distinguish between trusted and untrusted inputs.\n
- Capability inventory: Routes requests to various capabilities including $tool-connectors, which can perform authenticated API interactions with external marketing platforms.\n
- Sanitization: Incorporates 'adversarial checks' and 'claim ledgers' to verify information accuracy, providing a procedural layer of protection against misleading or malicious data.\n- [NO_CODE]: The skill consists entirely of markdown-based instructions and YAML configuration files. It does not include any executable scripts (Python, JavaScript, shell), minimizing the direct execution risk.
Audit Metadata