gke-manifest-generation
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- Indirect Prompt Injection: The skill generates Kubernetes manifests based on external inputs like natural language descriptions and code changes. This creates a potential surface where embedded instructions in that data could influence the agent's actions or the generated output.
- Ingestion points: The skill ingests natural language descriptions and application code changes (found in SKILL.md).
- Boundary markers: No specific delimiters or 'ignore instructions' warnings are defined for these ingestion points.
- Capability inventory: The skill is instructed to execute gcloud commands, perform file system writes, and utilize documentation search tools (found in SKILL.md).
- Sanitization: The instructions do not specify validation or sanitization routines for user-supplied parameters before they are used in shell commands, file paths, or incorporated into generated manifests.
- Command Execution: The skill directs the agent to interact with the gcloud CLI tool to generate optimized manifests. While this is a standard integration within the vendor's ecosystem, using shell commands with parameters derived from user input is a security consideration that warrants review.
- Evidence: The instruction to run gcloud container ai profiles manifests create with dynamic parameters (found in SKILL.md).
Audit Metadata