cloud-iam-deep
Pass
Audited by Gen Agent Trust Hub on Oct 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from target environments to identify cloud identifiers and credentials which are then used in agent commands.
- Ingestion points: File scanning for IdentityPoolId and credentials in references/details.md and SKILL.md.
- Boundary markers: Features a 'Mandatory confirmation gate' requiring the user to explicitly approve any command that probes or exploits a target.
- Capability inventory: Includes extensive subprocess execution via cloud CLI tools (aws, az, gcloud, kubectl) for reconnaissance and privilege escalation.
- Sanitization: Instructions advise on output redaction and the use of isolated CLI profiles for testing.
- [COMMAND_EXECUTION]: Contains numerous shell commands for interacting with cloud control planes. These operations are essential to the skill's stated purpose for security assessments and are mitigated by required human-in-the-loop confirmation.
Audit Metadata