refactor-legacy-assess

Pass

Audited by Gen Agent Trust Hub on Jul 6, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses local Node.js scripts for diagnostic checks, specifically referencing 'scripts/checklist.mjs' and a platform-specific diagnostic script at '~/.claude/skills/refactor-chain/scripts/diagnose.mjs'. These executions are restricted to internal utility scripts and do not involve untrusted input or remote code.\n- [PROMPT_INJECTION]: As the skill is designed to read and analyze external source code, it possesses a surface for indirect prompt injection where malicious instructions in the analyzed files could attempt to influence the agent's report output.\n
  • Ingestion points: Source code is read from the directory provided via the prompt (as described in SKILL.md).\n
  • Boundary markers: No explicit delimiters are used to wrap the analyzed code to prevent instructions within the code from being interpreted as agent commands.\n
  • Capability inventory: Subprocess execution of local Node.js scripts (SKILL.md) and local file read access.\n
  • Sanitization: No explicit sanitization or filtering of the analyzed code is performed prior to the assessment logic.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 6, 2026, 01:02 AM
Security Audit — agent-trust-hub — refactor-legacy-assess