ab-testing
Pass
Audited by Gen Agent Trust Hub on Nov 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect instructions through data ingestion and multi-step agent chaining.
- Ingestion points: The Initial Assessment section in SKILL.md directs the agent to read project-specific marketing files such as .agents/product-marketing.md or .claude/product-marketing.md.
- Boundary markers: The skill does not provide instructions to use delimiters or to ignore instructions embedded within these context files.
- Capability inventory: The skill's scope is limited to planning and analysis; no capabilities for shell command execution, file system modification, or network-based data exfiltration were found across the skill files.
- Sanitization: There is no evidence of sanitization or filtering of the content retrieved from the referenced context files.
- Multi-step chains: The skill is designed to cross-reference and defer tasks to other skills like copywriting or analytics, which could potentially propagate instructions from the ingested context.
Audit Metadata