optimize-skills

Warn

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • COMMAND_EXECUTION (MEDIUM): The file scripts/render-dot.py contains a relative path traversal string (../../../scripts/render-dot.py) instead of executable code. This pattern indicates an attempt to invoke a script from a parent directory, which may bypass directory-level isolation or execute unverified logic depending on the host environment's handling of such files.
  • REMOTE_CODE_EXECUTION (MEDIUM): The skill instructs the agent to execute scripts/render-dot.py to process DOT diagrams. Because the script's actual implementation is external to the skill package (referenced via path traversal), its safety cannot be verified during static analysis.
  • PROMPT_INJECTION (LOW): This skill is a utility for processing and optimizing other AI skills, which constitutes a surface for Indirect Prompt Injection (Category 8).
  • Ingestion points: The skill reads and audits SKILL.md files and .dot files from the local environment.
  • Boundary markers: There are no instructions provided to use XML delimiters or safety headers when the agent reads external skill files.
  • Capability inventory: The skill possesses the ability to execute scripts and write/modify files in the workspace.
  • Sanitization: The instructions lack specific guidance on sanitizing or escaping content found in the skills being optimized.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 17, 2026, 06:21 PM