evaluating-skills-with-models

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • PROMPT_INJECTION (LOW): The skill is susceptible to Indirect Prompt Injection due to how it processes test scenario files.
  • Ingestion points: The skill ingests untrusted data from tests/scenarios.md located within target skill directories.
  • Boundary markers: The prompt template used in Step 2 interpolates the {evaluation_query} directly without using delimiters or instructions to ignore embedded commands.
  • Capability inventory: The skill has the capability to spawn sub-agents that execute code, read files via Glob/Read, and modify local README files.
  • Sanitization: There is no evidence of sanitization or validation of the queries or expected behaviors parsed from the scenario files.
  • SAFE (SAFE): No instances of data exfiltration, hardcoded credentials, obfuscation, or persistence mechanisms were detected. The skill's functionality is consistent with its stated purpose of evaluating AI model performance.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:15 PM