ci-cd-and-automation
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes repository files and runtime inputs to generate CI/CD configurations, creating a surface for indirect prompt injection if the ingested data is maliciously crafted.
- Ingestion points: The workflow instructs the agent to inspect available files, runtime versions, and inputs within the local environment (SKILL.md).
- Boundary markers: The skill implements quality gates requiring the separation of recommendations from live changes and explicit user authorization before applying or deploying any configuration (SKILL.md).
- Capability inventory: The agent can produce implementation artifacts, run tests, and execute vendor-provided tools (sandbase_run) for documentation or inference (references/sandbase-api-map.md).
- Sanitization: The instructions mandate verifying outputs through tests, previews, and source comparisons rather than trusting file creation alone (SKILL.md).
- [SAFE]: The skill uses vendor-specific tools (sandbase_discover, sandbase_inspect, sandbase_run, etc.) for its intended purpose. These tools include built-in safeguards, such as requiring price estimates and user confirmation before execution, ensuring transparency and control over external interactions.
Audit Metadata