refactor-web-02-modules

Pass

Audited by Gen Agent Trust Hub on Jul 6, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes local shell commands using a script located at ~/.claude/skills/refactor-chain/scripts/orchestrate.mjs. This script is used for creating checkpoints (checkpoint --label pre-modules) and advancing the refactoring pipeline (advance --target <dir>). While these are local operations intended for state management, they represent a capability that interacts with the host environment.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its core functionality of processing untrusted data (the user's source code) and performing high-capability actions based on that processing.
  • Ingestion points: The skill inventories and audits every module directory and source file within the target project's feature layer.
  • Boundary markers: There are no explicit boundary markers or instructions to ignore embedded prompts within the processed source code.
  • Capability inventory: The skill can perform file moves, rewrites of import statements, and execution of shell scripts (orchestrate.mjs).
  • Sanitization: There is no evidence of sanitization or validation of the content within the user's files before they are processed by the refactoring logic.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 6, 2026, 01:02 AM
Security Audit — agent-trust-hub — refactor-web-02-modules