refactor-diagnose

Warn

Audited by Gen Agent Trust Hub on Jul 6, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONREMOTE_CODE_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes a Node.js script (diagnose.mjs) passing user input (utterance) as a command-line argument. This creates a potential command injection vulnerability if the input is not strictly sanitized by the execution environment.
  • [REMOTE_CODE_EXECUTION]: The core functionality is offloaded to a script located at ~/.claude/skills/refactor-chain/scripts/diagnose.mjs. Since this script is external to the skill's source files, its safety and behavior are unverifiable.
  • [DATA_EXFILTRATION]: The skill has the capability to write to the local filesystem via the diagnose.mjs learn command which modifies history.jsonl.
  • [PROMPT_INJECTION]: The skill's behavior (specifically the autopilot mode) is determined by classification logic that processes untrusted repository data, creating a surface for indirect prompt injection attacks where a malicious codebase could trick the agent into taking autonomous actions.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 6, 2026, 01:01 AM
Security Audit — agent-trust-hub — refactor-diagnose