refactor-companion
Pass
Audited by Gen Agent Trust Hub on Nov 1, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches design vocabulary, engineering principles, and composition guidelines from the author's public repository on GitHub (csark0812/toolbox).
- [DYNAMIC_EXECUTION]: Instructs the agent to generate and execute 'before/after' scripts that run entry points with various inputs to ensure refactoring does not change code behavior.
- [COMMAND_EXECUTION]: Uses local shell commands including 'git status', 'node', and 'tail' to inspect workspace state and verify changes.
- [INDIRECT_PROMPT_INJECTION]:
- Ingestion points: Reads and analyzes repository source code, configuration files, and test suites to identify callers and contracts (SKILL.md).
- Boundary markers: No explicit delimiters or instructions are provided to distinguish ingested code from core task instructions.
- Capability inventory: Includes the ability to perform file writes, script generation, and command execution across the local worktree.
- Sanitization: Does not implement sanitization or safety checks on repository content before it is used to influence refactoring logic or validation script generation.
Audit Metadata