verification-before-completion
Pass
Audited by Gen Agent Trust Hub on Nov 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow where the agent ingests and evaluates untrusted data—including code diffs, build artifacts, and execution logs—to verify completion claims. This ingestion creates a potential surface for indirect prompt injection if those external files contain instructions intended to divert the agent from its goals.\n
- Ingestion points: Repository diffs, build artifacts, and execution output (found in
SKILL.mdandreferences/completion-evidence.md).\n - Boundary markers: The instructions do not mandate the use of delimiters or safety warnings when processing this external content.\n
- Capability inventory: The skill guides the use of existing agent tools for file inspection and command execution but does not provide its own scripts.\n
- Sanitization: No specific data validation or sanitization steps are outlined for the inputs being verified.\n- [EXTERNAL_DOWNLOADS]: Mentions a design-drift workflow located in Google's official GitHub organization for reference purposes.
Audit Metadata