Protect AI / community · open-source
Rebuff
guardrail-testingprompt-testing
What it does
Prompt-injection detection library. Combines heuristics, vector-similarity to a corpus of known injections, and an LLM classifier; used both as a runtime defense and as a probe to evaluate defenses.
What it doesn't do
Not a comprehensive red-team harness — focused on PI detection. Like all pattern-and-classifier defenses, it can be bypassed by novel encodings; treat as defense in depth, not the sole gate.
Best for
- Adding a baseline PI-detection layer
- Generating evaluation prompts that target your existing defenses
Not for
- Replacing structured prompting and least-privilege tool authorization
Pricing
Free.
Integration effort
low
Team skill required
Basic Python.
Last reviewed: 2026-05-01