config-cost-optimization
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- Read-only AWS API Access: The skill is limited to read-only AWS operations, ensuring it cannot perform unauthorized state changes.
- Indirect Prompt Injection Surface: The skill processes potentially attacker-controlled resource metadata.
- Ingestion points: Ingests resource names, tags, and bucket identifiers via AWS APIs (SKILL.md Step 2).
- Boundary markers: Includes a 'Safety and Boundaries' section with explicit instructions to treat metadata as inert and ignore embedded directives.
- Capability inventory: Limited to read-only API calls across Config, S3, and Organizations services.
- Sanitization: Reliability is ensured via agent instructions and a manual validation loop (Step 5) instead of programmatic sanitization.
- Verified External Sources: References only official AWS documentation and vendor GitHub repositories.
- Secure Execution Context: No evidence of dynamic context injection or unauthorized shell execution patterns found.
Audit Metadata