amazon-ads-sb-performance-review
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes Sponsored Brands Campaigns Report data from external sources like CSV files, SQL databases, or API exports, creating a potential surface for indirect prompt injection. 1. Ingestion points: The SKILL.md workflow (Step 2) identifies the ingestion of external report data. 2. Boundary markers: The skill instructs the agent to state assumptions and confirm scope but does not define specific delimiters for the ingested data. 3. Capability inventory: The agent is authorized to generate and execute code for data loading and analysis as described in SKILL.md and references/agent-codegen-patterns.md. 4. Sanitization: While numeric safety is addressed (e.g., safe division), there is no explicit sanitization for potential prompt-injection payloads within the data headers or contents.
- [DYNAMIC_EXECUTION]: The skill explicitly directs the agent to author and run code at runtime to accommodate diverse data schemas and report formats. Evidence: SKILL.md requires the agent to 'write loader and column-mapping code' and 'write code to traverse keys' for JSON artifacts. references/agent-codegen-patterns.md provides templates for these dynamically generated scripts. This is categorized as low risk as it involves script generation from known templates for the skill's primary analytical purpose.
Audit Metadata