the-stockout-spend-guard
Pass
Audited by Gen Agent Trust Hub on Aug 14, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns or security risks were identified. The skill consists of instructional text for identifying inventory-spend mismatches and a reference guide for marketing operations.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external inventory and ad spend data, which is an ingestion surface for untrusted data. However, the risk is mitigated as the skill lacks high-risk capabilities like file writing, code execution, or network requests. It includes logic for data validation (grain matching and date freshness checks) and requires human intervention for all proposed actions.
- Ingestion points: Current inventory exports (SKU, on-hand units) and active ad spend reports as defined in SKILL.md.
- Boundary markers: Instructions require the agent to confirm data export dates and align data grains (variant vs. parent level) before processing.
- Capability inventory: None. The skill instructions explicitly state that the loop has no write access and must not apply changes autonomously.
- Sanitization: The process relies on explicit reporting of mismatches and date validation rather than automated data cleaning.
Audit Metadata