TokenFuzz 是 AI Skill Hub 本期精选AI工具之一。综合评分 7.5 分,整体质量较高。我们推荐使用将其纳入你的 AI 工具库,帮助提升工作效率。
TokenFuzz是开源AI工具,用于LLM(大语言模型)基于的漏洞研究平台,提供了一个开放的平台。
TokenFuzz 是一款基于 Shell 开发的开源工具,专注于 installable、shell 等核心功能。作为 GitHub 开源项目,它拥有活跃的社区支持和持续的版本迭代,代码完全透明可审计,支持本地部署以保护数据隐私。无论是个人使用还是集成到企业工作流,都能提供稳定可靠的解决方案。
TokenFuzz是开源AI工具,用于LLM(大语言模型)基于的漏洞研究平台,提供了一个开放的平台。
TokenFuzz 是一款基于 Shell 开发的开源工具,专注于 installable、shell 等核心功能。作为 GitHub 开源项目,它拥有活跃的社区支持和持续的版本迭代,代码完全透明可审计,支持本地部署以保护数据隐私。无论是个人使用还是集成到企业工作流,都能提供稳定可靠的解决方案。
# 克隆仓库 git clone https://github.com/tokenfuzz/tokenfuzz cd tokenfuzz # 查看安装说明 cat README.md # 按 README 完成环境依赖安装后即可使用
# 查看帮助 tokenfuzz --help # 基本运行 tokenfuzz [options] <input> # 详细使用说明请查阅文档 # https://github.com/tokenfuzz/tokenfuzz
# tokenfuzz 配置说明 # 查看配置选项 tokenfuzz --config-example > config.yml # 常见配置项 # output_dir: ./output # log_level: info # workers: 4 # 环境变量(覆盖配置文件) export TOKENFUZZ_CONFIG="/path/to/config.yml"
<p align="center"> <img src="docs/assets/logo-lockup.svg" alt="TokenFuzz" width="400"> </p>
TokenFuzz is an open platform for LLM-based vulnerability research: a coordinated fleet of agents that audits a codebase, finds security issues, and hands each one back with the evidence and reasoning a developer needs to triage it. Point it at any source tree you are authorized to test — C/C++, Rust, Go, Python, Java, and more, including browsers — and it runs as a pipeline:
- Recon sweep. A cold-start pass runs a CTF-style "find all vulnerabilities" survey, split across parallel agents over directory-coherent, LOC-balanced slices of the source. On a large codebase it scopes to recently-changed code so the pass stays bounded. Results land in a prioritized queue. - Eight investigation strategies. Deep agents work that queue with prior-fix mining, invariant negation, spec-vs-implementation, differential testing, lifetime and state, cross-project peer-fix mining, parser-input engineering, and property oracles — across Claude, Codex, Gemini, or a local Ollama model behind one probe-and-triage contract. - Reachability-labelled findings. Every finding separates attacker-controlled-byte issues from internal caller-misuse and pure test- or maintenance-tool surface, so triage moves on signal instead of drowning in null-derefs, OOMs, and assertion-only aborts. - Cost as a first-class resource. Prompt caching, capped state views, per-agent sanitizer-run budgets, soft turn caps, work-card leases, and SHA-pinned recon reuse keep unattended multi-agent runs affordable. - Fleet coordination. Shared logging and cluster-level dedup keep parallel agents accumulating work rather than repeating each other; an independent validator pass with no shared context catches a model's own reasoning errors before anything is accepted. - Maintainer-ready handoff. Every accepted crash exports as a bundle — sanitizer trace, reproducer testcase, one-command reproduce.sh, candidate fix direction, and optional patch.diff — that rebuilds against a clean upstream checkout.
The platform does the discovery, the analysis, the triage, and the handoff; the final security judgment stays with you.
TokenFuzz是一个开源AI工具,用于LLM基于的漏洞研究,提供了一个开放的平台,但其功能和使用场景需要进一步评估。
AI Skill Hub 为第三方内容聚合平台,本页面信息基于公开数据整理,不对工具功能和质量作任何法律背书。
建议在沙箱或测试环境中充分验证后,再部署至生产环境,并做好必要的安全评估。
✅ Apache 2.0 — 宽松开源协议,可商用,需保留版权声明和 NOTICE 文件,含专利授权条款。
经综合评估,TokenFuzz 在AI工具赛道中表现稳健,质量良好。如果你已有明确的使用需求,可以直接上手体验;如果还在评估阶段,建议对比同类工具后再做决策。
| 原始名称 | tokenfuzz |
| 原始描述 | 开源AI工具:TokenFuzz is an open platform for LLM-based vulnerability research。⭐9 · Shell |
| Topics | installableshell |
| GitHub | https://github.com/tokenfuzz/tokenfuzz |
| License | Apache-2.0 |
| 语言 | Shell |
收录时间:2026-05-30 · 更新时间:2026-06-02 · License:Apache-2.0 · AI Skill Hub 不对第三方内容的准确性作法律背书。