referrals
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFENO_CODEINDIRECT_PROMPT_INJECTION
Full Analysis
- [NO_CODE]: The skill consists exclusively of instructional documentation and strategic references, with no executable scripts, binaries, or command-running code present in any of the analyzed files.
- [INDIRECT_PROMPT_INJECTION]: The instructions establish a data ingestion surface by directing the agent to read marketing context from local workspace files like .agents/product-marketing.md.
- Ingestion points: .agents/product-marketing.md, .claude/product-marketing.md, product-marketing-context.md.
- Boundary markers: Not specified for the ingested content.
- Capability inventory: None (the skill itself has no executable script capabilities).
- Sanitization: No explicit validation or sanitization logic is defined for external context.
- [SAFE]: The skill does not contain hardcoded credentials, unauthorized network calls, or logic for data exfiltration. Tool recommendations are for industry-standard third-party services.
- [SAFE]: No evidence of obfuscation, metadata poisoning, or hidden malicious instructions was found during the analysis.
Audit Metadata