home/roleplay/real-prompt-injection-ab-eval

Real Prompt Injection ab Eval

GPTClaudeGemini··▲ 1,004 copies·updated 2026-07-14

real-prompt-injection-ab-eval.prompt

# Real Dataset A/B Prompt Robustness Evaluation

Source: [S-Labs/prompt-injection-dataset](https://huggingface.co/datasets/S-Labs/prompt-injection-dataset)

The suite evaluates two deterministic prompt-injection detectors on 320 public rows.

- v1: {'precision': 0.778, 'recall': 0.107, 'f1': 0.188}
- v2: {'precision': 0.635, 'recall': 0.252, 'f1': 0.361}

Interpretation: this is a real regression target. A production prompt classifier should improve
recall without simply overfitting to obvious keywords.

fill the variables

This prompt has 2 variables. Pro fills them into a ready-to-paste prompt for you — no manual find-and-replace.

{'precision': 0.778, 'recall': 0.107, 'f1': 0.188}{'precision': 0.635, 'recall': 0.252, 'f1': 0.361}

Unlock with Pro →

when to use it

Community prompt sourced from the open-source GitHub repo YutoTerashima/prompt-robustness-suite (MIT). A "Real Prompt Injection ab Eval" style prompt — adapt the placeholders and specifics to your task. Imported as-is and not independently retested here, so check the output before relying on it.