auditing-gcp-iam-permissions

Pass

Audited by Gen Agent Trust Hub on Apr 13, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill includes gcloud CLI commands that can modify cloud infrastructure state, specifically removing IAM policy bindings and deleting service account keys as part of remediation steps.
  • [DATA_EXPOSURE]: The skill accesses and enumerates sensitive GCP IAM configurations, including identity details and policy bindings, saving this information to a local JSON file (gcp_iam_audit.json).
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted metadata from GCP IAM policies and resource descriptions which could theoretically contain malicious instructions.
  • Ingestion points: Data is ingested through GCP API calls in scripts/agent.py and gcloud search results in SKILL.md.
  • Boundary markers: None present; data is directly processed into tables or JSON objects.
  • Capability inventory: Includes the ability to search assets, list service accounts, and modify IAM policies using gcloud commands in SKILL.md.
  • Sanitization: None; the skill relies on structured JSON output from GCP tools and standard Python data structures.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 13, 2026, 12:27 PM