verification-before-completion
Pass
Audited by Gen Agent Trust Hub on Aug 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [SAFE]: The skill implements a mandatory verification protocol for agents, prioritizing empirical evidence from tests and builds over natural language claims. This serves as a behavioral guardrail to prevent inaccuracies and ensure work integrity.- [INDIRECT_PROMPT_INJECTION]: The skill involves the ingestion of untrusted data from tool outputs (terminal commands, test reports, and VCS diffs). While this creates a surface for indirect prompt injection, the skill's specific purpose is to validate these outputs rather than blindly obey instructions embedded within them. Ingestion points include terminal execution output and agent reports, which are gated by a logical verification function that requires explicit confirmation before proceeding.- [COMMAND_EXECUTION]: The instructions reference executing terminal commands through an 'ExecutionProxy' interface (using the 'rtk' prefix). These operations are standard for development workflows, such as running linters, tests, and build tools, and are used here specifically for verification purposes.
Audit Metadata