refactor-assessment-map

Pass

Audited by Gen Agent Trust Hub on Jul 6, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses local shell commands to gather metrics and perform diagnostics. This includes running a Node.js script for signals and using 'git' to retrieve naming and churn data from the repository history.
  • Evidence: 'node ~/.claude/skills/refactor-chain/scripts/diagnose.mjs signals --target '
  • Evidence: 'git log --format= --name-only | sort | uniq -c'
  • [PROMPT_INJECTION]: An indirect prompt injection surface is present because the skill parses untrusted source code (imports/require statements) to construct its dependency map. This is a common attribute of code analysis tools.
  • Ingestion points: Project source code files as defined in 'SKILL.md' (Method step 2).
  • Boundary markers: Absent. The skill does not define specific delimiters to isolate parsed code content from the agent's instruction context.
  • Capability inventory: Shell execution of Node.js scripts and git commands; ability to rank targets for subsequent automated planning phases.
  • Sanitization: Absent. The skill performs static parsing but does not explicitly mention sanitization or validation of the extracted string content.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 6, 2026, 01:01 AM
Security Audit — agent-trust-hub — refactor-assessment-map