refactor-plugin

Pass

Audited by Gen Agent Trust Hub on Mar 29, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill operates as a high-level workflow controller. It validates the presence of a plugin by checking for local directory structures (skills/) or configuration files (.claude-plugin/plugin.json).
  • [SAFE]: Task execution and analysis are delegated to internal sub-skills (e.g., plugin-creator:assessor). There are no signs of hardcoded credentials, network exfiltration, or unauthorized command execution.
  • [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection because it reads and analyzes external source code provided via the <plugin-path> argument. However, the resulting capabilities are limited to writing local markdown plan files, which significantly reduces the risk.
  • Ingestion points: Files located within the user-provided <plugin_path> (SKILL.md).
  • Boundary markers: None explicitly defined for the ingested file content.
  • Capability inventory: File system read access and creation of plan files (via plugin-creator:assessor).
  • Sanitization: No specific sanitization or filtering of the ingested source code is described in the workflow instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 29, 2026, 08:41 AM
Security Audit — agent-trust-hub — refactor-plugin