atomise

Pass

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: LOWPROMPT_INJECTIONNO_CODE
Full Analysis
  • [PROMPT_INJECTION] (LOW): The skill is vulnerable to indirect prompt injection because it processes untrusted user input within its reasoning loop. 1. Ingestion points: User-provided text enters the agent context through the <problem> parameter in the /atomise command. 2. Boundary markers: The skill instructions use double-quote delimiters for the input. 3. Capability inventory: The skill is restricted to 'Read' operations, limiting the impact to the agent's internal reasoning and local state. 4. Sanitization: No sanitization or filtering of the input content is specified in the framework.
  • [NO_CODE] (INFO): This skill consists entirely of Markdown instructions and does not include any scripts, executables, or external packages, which eliminates risks associated with remote code execution or malicious dependencies.
Audit Metadata
Risk Level
LOW
Analyzed
Feb 16, 2026, 08:53 AM