file-organizer-zh-hardened
Pass
Audited by Gen Agent Trust Hub on Apr 21, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill provides an attack surface for indirect prompt injection because it reads and processes untrusted file names and directory listings which could contain malicious instructions for the agent.
- Ingestion points: The skill scans local directories and reads file names using
fs.readdirSyncinindex.js. - Boundary markers: The
SKILL.mdfile contains explicit instructions to the agent to present a plan and wait for user confirmation before performing destructive operations. - Capability inventory: The skill has permissions to read directory contents (
fs.readdirSync) and move or rename files (fs.renameSync). - Sanitization: The code uses Node.js
pathutilities for path assembly and normalization but does not perform content-based sanitization of file name strings before they are presented in the agent's context.
Audit Metadata