skills/ian-pascoe/dotfiles/simplify/Gen Agent Trust Hub

simplify

Pass

Audited by Gen Agent Trust Hub on Mar 29, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: The skill creates an indirect prompt injection surface by instructing the agent to ingest and rewrite untrusted data (recently modified code). 1. Ingestion points: Recently modified code sections. 2. Boundary markers: None present. 3. Capability inventory: Implicit write access to refine and simplify code files. 4. Sanitization: None present. Additionally, the instruction to operate autonomously 'without requiring explicit requests' reduces user oversight and may allow the agent to process malicious instructions embedded in modified code.
  • [EXTERNAL_DOWNLOADS]: The skill references 'http://CLAUDE.md' for coding standards. While this likely refers to a local project file following platform conventions, it is syntactically a URL pointing to a non-whitelisted host.
  • [SAFE]: The core instructions focus on improving code readability and maintainability without altering functionality, which are standard software engineering practices.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 29, 2026, 03:29 AM
Security Audit — agent-trust-hub — simplify