refactor-backend-04-service

Pass

Audited by Gen Agent Trust Hub on Jul 6, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes a local utility script scripts/checklist.mjs using Node.js to output its operational steps in JSON format for the agent's internal use.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes untrusted source code from external projects.
  • Ingestion points: Reads .java, .xml, .yml, .yaml, and .properties files from the target repository to classify services and update references.
  • Boundary markers: The skill implements a strict confirmation-gated flow, requiring the user to explicitly approve a 'Fix Plan' before any modifications occur. It also utilizes a four-way verification phase after execution.
  • Capability inventory: Performs file system writes, file moves, and deletion of empty directories. It executes workspace-wide search-and-replace for package names and imports.
  • Sanitization: The instructions contain strict constraints to only modify package and import lines, explicitly prohibiting changes to method bodies, annotations, or comments, which limits the impact of potentially malicious content in the processed files.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 6, 2026, 01:01 AM
Security Audit — agent-trust-hub — refactor-backend-04-service