Metadata-Version: 2.4
Name: superred-optimizer-crescendo
Version: 0.1.0
Summary: Crescendo multi-turn jailbreak optimizer for superred
Project-URL: Homepage, https://superred.simonsure.com
Project-URL: Documentation, https://superred.simonsure.com
Project-URL: Repository, https://github.com/RoldSI/superred-modules
Project-URL: Issues, https://github.com/RoldSI/superred-modules/issues
Author-email: Simon Sure <info@simonsure.com>, Sarthak Munshi <sarthakmunshi@gmail.com>
License-Expression: MIT
License-File: LICENSE
License-File: LICENSES/NOTICE.md
License-File: LICENSES/PyRIT-MIT.txt
License-File: NOTICE
Keywords: ai-safety,crescendo,jailbreak,llm-security,multi-turn,red-teaming,superred
Classifier: Development Status :: 3 - Alpha
Classifier: Intended Audience :: Science/Research
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.11
Classifier: Programming Language :: Python :: 3.12
Classifier: Programming Language :: Python :: 3.13
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
Classifier: Topic :: Security
Requires-Python: <3.14,>=3.11
Requires-Dist: superred>=0.1.0
Provides-Extra: test
Requires-Dist: pytest; extra == 'test'
Requires-Dist: pytest-asyncio; extra == 'test'
Description-Content-Type: text/markdown

# superred-optimizer-crescendo

Crescendo multi-turn jailbreak optimizer for the
[superred](https://superred.simonsure.com) red-teaming framework.

Crescendo escalates a conversation from benign to harmful over several turns,
backtracking and restoring history when the target refuses, until the
conversation objective is achieved.

## Install

```bash
pip install superred-optimizer-crescendo
```

## Credits / upstream

This optimizer reimplements the **Crescendo** multi-turn jailbreak
(Russinovich, Salem, Eldan; USENIX Security 2025 —
https://crescendo-the-multiturn-jailbreak.github.io/).

The bundled attacker prompt variants
(`src/crescendo_optimizer/prompts/variant_1..5.py`) and the task-achieved
scoring prompt (`evaluator.py`) are taken from **Microsoft PyRIT**
(https://github.com/microsoft/PyRIT), MIT licensed, Copyright (c) Microsoft
Corporation. See `LICENSES/NOTICE.md` for per-file attribution.

Our optimizer harness, evaluator logic, capability-aware extensions, and
deterministic-replay mechanism are original superred code, MIT licensed
(see `LICENSE`).
