validate-consistency
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted project files (Markdown, YAML, source code) to perform consistency checks. This represents an inherent attack surface where malicious instructions embedded in project files could attempt to influence the agent's behavior.\n
- Ingestion points: Scans all project documentation (README, CLAUDE.md, CHANGELOG, etc.), source code, and CI workflow YAML files via the Read and Glob tools.\n
- Boundary markers: The instructions do not specify the use of delimiters or warnings to ignore embedded instructions within processed data.\n
- Capability inventory: Tool use is restricted to Read, Glob, Grep, and specific Bash subcommands (echo, git, diff). It cannot perform network operations or execute arbitrary shell commands.\n
- Sanitization: There is no evidence of content sanitization, escaping, or validation of the external documentation content before interpolation into prompts.
Audit Metadata