skills/niller2005/polyflup/trainer/Gen Agent Trust Hub

trainer

Fail

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: HIGHPROMPT_INJECTION
Full Analysis
  • Indirect Prompt Injection (HIGH): The skill is designed to read external, potentially untrusted content (commits and code) and use it to modify the system's instruction set (SKILL.md files).
  • Ingestion points: Workflow step 2 involves reading commits and code changes from the repository.
  • Boundary markers: None identified; the skill directly translates code observations into skill updates without delimiters or 'ignore' instructions.
  • Capability inventory: Workflow step 3 grants the skill the ability to 'Edit or create SKILL.md files', which defines the behavior of AI agents, and manage 'AGENTS.md'.
  • Sanitization: No validation or human-in-the-loop review process is mentioned for updating instruction files, allowing malicious code to potentially redefine system constraints.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 16, 2026, 10:27 AM