gcp-secret-manager
Pass
Audited by Gen Agent Trust Hub on Oct 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes externally controlled data such as secret names, IAM policy members, and Cloud Audit Log entries. * Ingestion points: SKILL.md (via gcloud secrets list, gcloud secrets get-iam-policy, and gcloud logging read commands). * Boundary markers: The instructions do not define specific delimiters to separate untrusted data from agent instructions. * Capability inventory: The skill executes shell commands using gcloud, jq, and bash built-ins for resource discovery and reporting. * Sanitization: Data is parsed using jq to extract specific fields, which provides structural isolation but does not filter content for malicious instructions.
- [COMMAND_EXECUTION]: The skill relies on executing local shell commands to interact with Google Cloud services. * Evidence: Multiple bash scripts in SKILL.md utilize gcloud, jq, xargs, and basename for resource discovery and automated analysis.
Audit Metadata