aegis-cloud-iam
Pass
Audited by Gen Agent Trust Hub on Aug 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted Infrastructure-as-Code (IaC) files and repository artifacts, which presents a surface for indirect prompt injection. Malicious instructions could potentially be embedded in the metadata or comments of the files being audited to influence the agent's findings.
- Ingestion points: Audited repository files including Terraform (.tf), CloudFormation (.yaml/.json), and synthesized templates (SKILL.md).
- Boundary markers: The instructions include a 'Needs-Review' status marker to handle runtime facts that cannot be proven from static analysis, though it does not define explicit delimiters for untrusted input data (SKILL.md).
- Capability inventory: The skill involves complex policy reasoning and string-based pattern matching (via ripgrep) across file contents (SKILL.md).
- Sanitization: The instructions do not specify explicit sanitization or filtering mechanisms for the input data.
- [EXTERNAL_DOWNLOADS]: The skill references security guidelines, standards, and vendor documentation from trusted sources and well-known services, including Amazon Web Services, Microsoft Azure, Google Cloud Platform, GitHub, and OWASP. These references are used to provide technical context for IAM audit findings and do not involve the execution of untrusted remote code (further-reading.md).
Audit Metadata