review-changes
Pass
Audited by Gen Agent Trust Hub on Nov 5, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes repository scripts such as 'npm run verify', 'npm run review:worktree', and 'npm ci' to create isolated environments and validate code changes. These commands run locally and are part of the standard development workflow for this repository.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests external data in the form of git diffs and untracked files which are then processed by subagents, potentially allowing malicious instructions in the code to influence the agents. 1. Ingestion points: Git diffs and untracked files are supplied to reviewers and Doistbot. 2. Boundary markers: The reviewer brief template lacks specific delimiters or warnings to isolate untrusted content from the subagent's instructions. 3. Capability inventory: Reviewing agents can execute shell commands and interact with the doistbot CLI tool. 4. Sanitization: The instructions advise the user to manually check untracked files for credentials and personal data before sharing them with the agents.
Audit Metadata