安全项目
正在加载…
安全项目
在代码仓库中发现AI编程工具痕迹的公开项目，按最近收录排序。
安全 · 1,336个项目清除筛选
One read-only query in your Supabase SQL editor. Every open policy, table and function, quoted from your database, with the fix and the undo. Free.
翻译准备中
Prismor intercepts every tool call your AI agents make, enforces security policies, and logs a full audit trail. Works with all enterprise AI agents.
翻译准备中
SIQ Agent Security：面向 Agent Skills 的可信执行安全运行时。Skill 引导接入、本地 Go 程序裁决、宿主钩子执行保护，以 Ed25519 与 SHA-256 绑定授权、参数与回执；0.4.0 签名正式版已发布。
翻译准备中
La preuve qualifiée eIDAS, ouverte, automatisable et protégée par des statuts intangibles. Un pilote tourne déjà. Livre blanc en consultation, appel à partenaires.
翻译准备中
한국어 개인정보 비식별화 오픈소스 엔진 — 규칙+로컬 LLM 하이브리드 PII 탐지·마스킹 라이브러리, CLI, MCP 서버 (2026 오픈소스 개발자대회, 팀 마스킹테이프)
翻译准备中
检查您的电子邮件是否在数据泄露中被曝光，并获取 AI 驱动的风险评分以及明确的降低曝光的步骤。
AI 翻译
Signed receipts for vibe-coded data pipelines. Proves nobody changed your data, and shows the exact link if they did.
翻译准备中
Vero is open-source zero-knowledge compliance for AI agent payments. Agents prove eligibility — not identity. MIT licensed. FATF R.15/R.16 aligned. Built on the Open Wallet Standard. Fork it on GitHub
翻译准备中
Open-source privacy and approval gateway for AI assistants (MCP): human approval, local PII checks and audit for Gmail, Drive, Slack, Salesforce, Jira and more.
翻译准备中
CyberArena is a live benchmark where two AI agents fight an attack-and-defense CTF: each one exploits the other's vulnerable service while patching its own, round after round. Watch live matches, repl
翻译准备中
CRANE — self-hosted compliance management for the EU Cyber Resilience Act. SBOM analysis, vulnerability tracking, release gates and audit trails in one auditable platform.
翻译准备中
Browse live police cameras on a map, filter by department, and build a video wall
翻译准备中
AI 摘要 · 网站与 README
Hikyo is a fully open-source, self-hosted control plane that makes configuration state explicit across every environment.
翻译准备中
Write a team's threat model as text files in git, compiled and checked automatically
翻译准备中
AI 摘要 · 审核时修订
AI cloud video management for the cameras you already own: live streaming, 23 detections and scheduled virtual patrol rounds with a compliance record.
翻译准备中
Task-scoped warrants that narrow as agents delegate. Every hop signed, every chain verifiable offline.
翻译准备中
Streamline incident management with a comprehensive frontend that offers intuitive dashboards, investigation workflows, and real-time notifications for efficient response.
翻译准备中
KeePassXC Material for Windows: downloads, documentation, and release provenance.
翻译准备中
Open-source customer due diligence risk checks for any legal entity. Powered by GLEIF, OpenSanctions, national company registries and the Beneficial Ownership Data Standard.
翻译准备中
Diagnóstico e correção de vulnerabilidades em sistemas web para PMEs e fintechs. OWASP Top 10 · TLS · LGPD art. 46. Suíte de ferramentas open source.
翻译准备中
Open-source credential broker that issues short-lived SSH, AWS, GitHub, and Kubernetes credentials after FIDO2 hardware verification.
翻译准备中
SOVP is a deterministic AI infrastructure certification standard: 180+ parameters, a three-state verdict, Ed25519-signed. U.S. Patent Pending.
翻译准备中
Your scanners find issues. This finds the way in: the route from internet exposure, through too much privilege, to a sensitive asset - caught in the pull request that opens it.
翻译准备中
Esver OS turns repeated company workflows into verified executable skills. Visage keeps identity local. Mr. Haven keeps capital programmable.
翻译准备中
Your .env files hold only references. The real values live in an encrypted vault on your machine. No account, no cloud, no telemetry.
翻译准备中
Écosystème open source pour reprendre la main sur son droit, son identité, sa sécurité et son économie.
翻译准备中
Approval gates and runtime authorization for AI coding agents. Decide what agents such as Claude Code, Codex and Cursor may do, what is denied, and what requires human approval — decided before integr
翻译准备中
简短、供应商中立的指南，介绍硬件安全密钥的实际功能——FIDO2 通行码、TOTP、智能卡（PIV）和 OpenPGP——如何设置以及在哪里阅读正式标准。
AI 翻译
vet402 buys what x402 endpoints actually sell, verifies fulfillment against the seller's own declaration, and publishes the results with evidence.
翻译准备中
A self-hosted OpenID Connect provider with a bundled OpenLDAP directory and a web management UI. One login for your modern apps, one LDAP directory for the rest, no phone-home.
翻译准备中
A sanctions-screening audit: which way of comparing names, and where you set the bar, measured on your own alert history.
翻译准备中
Slips 是一套自由软件、基于行为的 Python 入侵防御系统（IDS/IPS），使用机器学习检测网络流量中的恶意行为。Stratosphere Laboratory、AIC、FEL、CVUT
AI 翻译
A fruit-fly escape reflex, compiled to 173 NAND gates and 6 latches, evaluated gate by gate in your browser inside a voxel world.
翻译准备中
A unified, one-command SSO Manager + OIDC proxy stack for home labs and small businesses. Wires together a self-hosted identity provider and a reverse proxy with one setup.sh.
翻译准备中
Free read-only Lite lists every custom Wazuh rule that never fired and the likely reason, including rules dropped at load (7617/7619).
翻译准备中
A web management tool that supplements the Wazuh Dashboard: agents, groups, cluster nodes, rules and decoders, ready-made detection packs, API users and logs. Bilingual English / 繁體中文, AGPL-3.0.
翻译准备中
Your infrastructure, off the map. The qURL Connector hides your apps and APIs from public scans. One API call creates a private access link for a person or AI agent, with the scope, expiry, and sessio
翻译准备中
ThreatLens simulates realistic SOC investigations across identity, endpoint, email, and cloud — and grades your investigation against a hidden ground truth. Not a SIEM. A cyber range for analysts.
翻译准备中
Evidence-first network configuration auditing: every authoritative finding cites the configuration line that proves it.
翻译准备中
Check an OpenCloud instance for known vulnerabilities, missing hardening, weak security headers and available updates. Free to use, with no account required.
翻译准备中
GameDay testing for MCP servers. Four named experiments over the JSON-RPC wire, in a sandbox, with scripted and local-model seats. Receipts, tapes and a dataset come out. The oracle is the wire, not t
翻译准备中
Holdcall sits between your agent and its MCP tools. It decides every call on your machine, shows you the real arguments before anything dangerous runs, and keeps a record you can verify offline.
翻译准备中
A small key that plugs into USB. Touch it to unlock, approve and authenticate — never a password.
翻译准备中
Every connection across the organisation, wherever it runs, protected against tomorrow's decryption. Control stays in-house, always.
翻译准备中
Get real SIM-based temporary phone numbers instantly. Receive SMS verification codes for 1,300+ services across 150+ countries. Pay only when a code arrives — automatic refund if it doesn't.
翻译准备中
Arcjet is the AI agent runtime security platform. Discover the agents running in your organization, enforce policy across every action, prompt, and tool call, and keep the evidence to prove what happe
翻译准备中
Manage your Keycloak from your AI assistant: clusters, realms, apps, SSO, custom domains, WAF, webhooks and SIEM. One command, browser sign-in, no API key.
翻译准备中
Detection rules (Sigma-compatible core), real-time Endpoint Security, behavioural scoring, AI coding guardrails, and Agent Traces — W3C TRACEPARENT correlation between AI-agent activity and kernel eve
翻译准备中
Agadar — scam-message detection for Sorani Kurdish and Iraqi Arabic, with the reasoning shown. Runs entirely in your browser. Python, scikit-learn, char n-grams.
翻译准备中
An open standard letting a nonprofit publish, from its own domain, a machine-readable declaration of the giving destinations it authorizes. Consumed by AI agents through W3C WebMCP.
翻译准备中
Działa na Twoim komputerze. 98,9% pełnych nazwisk, 23 rodzaje numerów, podmiany odwracalne.
翻译准备中
See where any domain stands in the BlackVeil Global Security Index, or run 17 NIST-aligned checks free — then verify, fix and continuously monitor it.
翻译准备中
Cognitive Deterministic Memory Security OS for Agentic AI that reaches backward through time to find the quiet change behind today's failure, not the lookalike. Intelligence is not a cloud subscriptio
翻译准备中
OpenAI 的 Codex Security CLI 和 TypeScript SDK，用于发现、验证和修复安全漏洞。npm: https://www.npmjs.com/package/@openai/codex-security
AI 翻译
面向 AI 代理与 APIs 的执行授权网关。它拦截有后果的 HTTP 动作，询问 Decionis 它们是否已获授权，并恰好把已授权的请求转发一次，在单个
AI 翻译
pinact 是一个用于编辑 GitHub Workflow 与 Composite action 文件，并固定 Actions 与 Reusable Workflows 版本的 CLI。pinact 也可以更新这些版本并校验版本注释。
AI 翻译
Single-file browser client for the Have I Been Pwned API. No API key, no backend. Passwords are hashed locally and only a 5-character SHA-1 prefix is ever sent.
翻译准备中
Recover ZecWallet Lite seeds, zcashd wallet.dat files, and standalone Sapling or Sprout keys with a local-first desktop flow.
翻译准备中
Free sanctions screening, PEP checks, and adverse media search tool. Plus a comprehensive AML Knowledge Hub covering MLR 2017, POCA 2002, and FCA compliance. Built for compliance professionals.
翻译准备中
Something feels off? Check the vibe. A second opinion for messages, profiles, invitations and offers: warning signs of scams, grooming, coercion and exploitation, and where to get help.
翻译准备中
Live in-browser multi-person pose tracking and zone-based behavior detection. Runs entirely client-side, no backend, no upload.
翻译准备中
106 次精选的开源 AI 代理、MCP 服务器和 LLM 应用安全扫描——24 项已确认修复，本地优先运行，使用 Semgrep、Gitleaks、Trivy 和 pip-audit。
AI 翻译
The Kubernetes-native policy enforcement plane for MCP. Deterministic admission and egress policy, attributable audit, and SIEM export — MIT, self-hosted, no SaaS.
翻译准备中
Sandboxes isolados e descartáveis para agentes de IA: MCP, CLI, dashboard web e app nativo de macOS. Rede desligada por padrão, allowlist de domínios e trilha de auditoria. Go + SwiftUI.
翻译准备中
Replays sandbox test results comparing how different Git workflow setups resist attacks
翻译准备中
AI 摘要 · 审核时修订
Face ID-style unlock for Omarchy: the infrared camera recognises you at the lock screen, at sudo and in polkit, while your password stays the real key.
翻译准备中
Open-source Aadhaar, PAN, GSTIN & UPI detection and reversible masking for AI prompts. Runs on your infrastructure. MIT.
翻译准备中
The primary threat is accidental destruction — an agent that misreads a path and deletes the wrong folder — not a determined attacker. Network isolation is secondary and off by default; an AI CLI that
翻译准备中
CI/lock collects trusted telemetry across every checkpoint between commit and production — CI/CD builds, security scans (SAST/DAST/SBOM), runtime + cluster integrity (Falco, Linkerd mTLS, kube-bench),
翻译准备中
The local-first credential vault your AI agents can use — but never see. 66 testable security constraints. Rust.
翻译准备中
AI Agent City is the operations layer for the agent economy: spending limits, monitoring, security, trust, and discovery for AI agents. Human-readable and machine-readable.
翻译准备中
GDPR Chapter V Transfer Impact Assessment skill for Claude — EDPB six-step methodology, CNIL three-way Step 3, four essential guarantees, 12 country profiles, balanced Art. 49 analysis, RoPA interchan
翻译准备中
GDPR Breach Response Sentinel Agent Skill (open standard) — ENISA severity assessment, EDPB case matching, Art. 33/34 notification support, audit-ready .docx generation
翻译准备中
Read-only macOS notch + menu bar monitor for Claude Code and Codex. Open source, free, local-only. Risk-scores every tool call and tracks the footholds agents leave behind.
翻译准备中
A policy-enforcing filesystem for AI agents. Point it at a real project directory and it serves a filtered mirror — same paths, same sizes — deciding per read which face of each file to show.
翻译准备中
Low-latency bot detection for Apache 2.4 with proof-of-work challenges, shared-memory reputation, crawler policy, and built-in observability.
翻译准备中
Short-lived SSH certificates from your identity provider, for ssh, sudo, and the console.
翻译准备中
mcp-skills-vault installs MCP servers safely. mcp-trace shows what they're actually doing. Offline-first. No telemetry.
翻译准备中
Trust intelligence scoring for OSINT sources — behavioural reliability analysis, GDPR & EU AI Act compliant.
翻译准备中
Self-hosted MCP gateway: AI agents get audited, SSO-gated, read-only database access without ever holding a DB credential.
翻译准备中
SWARM: open-source multi-agent AI safety framework. Measure toxicity, governance gaps, and emergent risks with soft probabilistic labels.
翻译准备中
Belay is a local-first security layer at the tool-call boundary of AI coding agents. Deterministic, sub-100ms, fail-closed. Detect, block, and get notified — no LLM in the decision path.
翻译准备中
Hub + agent + cihaz entegrasyonları: canlı paket ölçümü, süreç bazlı L7/DNS görünürlüğü, TimescaleDB + NATS ölçek altyapısı, RBAC/SSO, 5651 uyumlu imzalı loglar. Açık kaynak, MIT, kendi altyapınızda.
翻译准备中
Production-ready security scanning platform with 28 scanners, OPA Policy-as-Code, historical storage, trend analysis, and compliance automation across 6 asset types
翻译准备中
A deny-by-default contract layer for AI agent tool calls - strict argument validation, ghost-argument stripping and self-healing retries, in-process and Pydantic-only
翻译准备中
Govern connected AI clients, models, and tools on your infrastructure. Centralize access policies and audit, then build custom capabilities on the SystemPrompt runtime.
翻译准备中
Free, open-source CLI and MCP server that keeps API keys in your OS keychain and serves them to Cursor, Claude Code, and Kiro. No cloud, no account, no .env file.
翻译准备中
AI security from risk map to reference architecture: the full CoSAI taxonomy, the MITRE mitigations that support its controls, and target-state architectures for every class of AI application.
翻译准备中
177 scenarios across 16 Cortex detection planes. Controlled signal generation for XSIAM/XDR detection validation.
翻译准备中
A rubber band around Claude's claws. Blocks rm -rf and other destructive commands before they run, so bypassPermissions mode is actually safe.
翻译准备中
Collective discovery for open cryptographic challenges. Contribute GPU time to a public elliptic-curve search, earn credit for verified work, share the prize when the answer is found.
翻译准备中
Choose the assets. Run applicable upstream security checks. Read one report organized by asset and priority.
翻译准备中
Test your vibe-coded apps, hunt down hallucinated packages, logic bugs, and secret leaks, and fix them in seconds. Works inside Cursor, Claude Code, GitHub Copilot, and Windsurf.
翻译准备中
Small Print is a nightly record of the tool descriptions, schemas and instructions that MCP servers, agent skills and plugins give your agent: every version hashed, every change diffed and graded by a
翻译准备中
Sane WordPress defaults with actionable Site Health reporting for security, privacy, performance, and core updates — including same-line security patches.
翻译准备中
The archive feature Graylog Open is missing — self-hosted, streaming, with independent compliance audit.
翻译准备中
AI Ops for Bring Your Own Agent (BYOA) teams. Audit coworker harnesses with doorman dashboard, mine prompt history for capability needs, and block rogue tools at the gate.
翻译准备中
Install Know Your Agent with npx in minutes. Offline demo, MCP gate, local free console, or enterprise hosted. No private protocol.
翻译准备中
Daily Internet.nl standards measurements of a real fleet, made by the unmodified internetnl CLI against a self-hosted netnl facade.
翻译准备中
Your agent knows a lot about you. Sanctuary keeps that information private. A firewall for your AI agent that runs on your own Mac, enforcing what it can reach at the operating-system level with keys
翻译准备中
做出来了，现在该展示了。
在项目文件夹中执行一条命令即可。只要由AI辅助制作，就不必具备AI功能。
/nomorevibe查看收录流程