Live Tool · Open Source
System Prompt Auditor
Paste an LLM system prompt or config to grade it against defensive best practices — secret isolation, instruction/data separation, least-privilege tool use, refusal policy, and more.
100% client-side · nothing is uploaded
0 chars
Try a sample
Your report card will appear here — a grade, per-check results, and prioritized fixes.
What it checks
The auditor runs a transparent set of heuristics inspired by the OWASP Top 10 for LLMs. It is a guide, not a guarantee — pair a good grade with runtime defenses, output validation, and testing.
Secret isolation
Flags API keys, tokens, and credentials that should never live in a prompt.
Injection resistance
Checks for instruction/data separation and rules that resist override and extraction.
Least-privilege agency
When tools are referenced, checks for allow-lists and approval gates on risky actions.