current-file-review

Pass

Audited by Gen Agent Trust Hub on Apr 23, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [SAFE]: The skill operates in a 'plan-before-execute' mode, ensuring that the AI agent cannot modify files without direct human authorization.
  • [COMMAND_EXECUTION]: Uses standard utilities (cat, git diff) to access project context. These commands are localized to the workspace and represent standard developer workflows.
  • [PROMPT_INJECTION]: The skill processes untrusted input from project files and code changes. Ingestion points: Project guidelines (CLAUDE.md, AGENT.md) and source code diffs. Boundary markers: Results are presented with distinct markdown headers. Capability inventory: File reading and writing capabilities. Sanitization: Mitigation is handled via a mandatory user approval step for all suggested modifications.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 23, 2026, 09:48 AM