devops-tooling
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill documents patterns for fetching and processing data from external endpoints (e.g., using
curlandInvoke-RestMethodinSKILL.mdandreferences/shell-scripting-patterns.md). This creates an attack surface for tool output poisoning where an attacker-controlled remote response could embed instructions to influence the agent's behavior. - Ingestion points:
SKILL.md(Part 2 and Part 3),references/shell-scripting-patterns.md(JSON/YAML processing sections). - Boundary markers: Absent in code examples that demonstrate data interpolation.
- Capability inventory: The skill enables network operations (
curl,wget,Invoke-RestMethod) and file system access across multiple scripts. - Sanitization: The skill provides generic guidance in the "DevOps Best Practices" section to "validate inputs and parameters," but lacks specific implementation examples for sanitizing remote data.
- [EXTERNAL_DOWNLOADS]: The skill references several well-known and trusted external tools and GitHub Actions for CI/CD workflows (e.g.,
actions/checkout,hashicorp/setup-terraform,codecov/codecov-action). These references are consistent with standard DevOps practices and originate from established providers.
Audit Metadata