changes-review
Pass
Audited by Gen Agent Trust Hub on Nov 4, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes
gitcommands (e.g.,git status,git diff) and local Node.js scripts such as.claude/scripts/ai-signal-scan.cjsand.claude/hooks/lib/review-receipt.cjsto manage the review workflow and perform analysis. These are part of the author's provided infrastructure for the skill. - [DYNAMIC_EXECUTION]: The skill runs Node.js scripts located within the project's
.claudedirectory. These tools are used for specialized tasks like scanning for AI-related signals and managing review snapshots during the fix loop. - [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted data from source code changes (diffs). Since the agent has the capability to write to the filesystem and execute local commands, malicious instructions embedded in the reviewed code could potentially attempt to influence the agent's behavior.
- Ingestion points: Git status output, git diff output, and targeted file contents as defined in
SKILL.md. - Boundary markers: None explicitly specified in the instructions for the diff content processed by the agent.
- Capability inventory: Execution of shell commands (git, node) and file-write access for applying automated fixes.
- Sanitization: The instructions do not specify any sanitization or validation of the ingested code content before processing.
Audit Metadata