Veja a ordem dos plugins no diretório público do ChatGPT por data de captura. Popular é uma seleção editorial. O diretório não divulga números de instalações ou de uso.
Última observação
The Doers Firm
Cloudflare Security reviews Cloudflare account configuration, Workers, Pages, D1, storage, APIs, DNS, TLS, origins, WAF, and deployment workflows. It uses authorized read-only account data and supplied code to produce evidence-backed findings, prioritized fixes, side-effect warnings, and verification steps. Customer support: https://thedoersfirm.com/support. It cannot guarantee complete security.
PolicyLayer
Check configured or proposed MCP servers against PolicyLayer registry records for identity, authentication posture, tool capabilities, risk grade and recent changes.
RelayShield
RelayShield Free Scam Checks screens links, crypto wallets, suspicious emails, and email addresses against RelayShield's threat-intelligence corpus — 115 monitored Telegram marketplaces, 494K+ indicators, 7.8M+ citations — before you click, pay, or reply. Four read-only tools: check a URL for phishing and malware, screen a crypto wallet for sanctions and scam flags, score a suspicious email for phishing signals, and look up whether an email appears in known data breaches. Free, no signup, no API key. Verdicts are FLAGGED or unknown — the tools never report anything as safe, because absence of evidence is not evidence of absence.
Safeguard
Safeguard helps users inspect software supply chain risk from ChatGPT, including vulnerabilities, SBOMs, findings, projects, components, policies, guardrails, compliance reports, remediation plans, SCM integrations, notifications, and organization workflows.
DUC THANG LUU
Inspect repository policy, GitHub Actions trust boundaries, and moderation automation without network access or repository mutation; generate reviewable remediation patches on request.
Audit real security risks
Endor Labs
Packaged AppSec workflows for SCA remediation, AI SAST triage, CI/CD posture, malware response, findings review, Endor Labs setup, and more.
The Doers Firm
Review prompts, AI application designs, retrieved documents, web content, and tool workflows for prompt-injection risks. Get evidence-linked attack-path analysis, calibrated severity and confidence, layered mitigations, and safe regression tests. Results are advisory and cannot guarantee prevention. Customer support: https://thedoersfirm.com/support.
Evidence-backed app security
Minimus
Minimus plugin that teaches AI coding tools to build Dockerfiles and Kubernetes manifests with hardened near-zero-CVE images
Aivana
Translates user requests into transparent KQL and executes bounded, read-only Defender XDR queries through a direct OAuth/OBO API, then produces reproducible evidence packages and analyst review without response execution.
Issam Chaaban
An independent skills-only workflow for authorized security research. Uses tools already connected to the active chat. Supports source review, scoped web/API investigation, validation, and HackerOne report drafting. Does not bundle testing tools or the Codex Security backend.
NightVision Security, Inc.
Guide NightVision API discovery, DAST scan configuration, CI/CD integration, and security-finding triage using the local NightVision CLI.
Guardio
Guardio helps you check suspicious links and understand your online protection from ChatGPT and Codex. Check a website before opening it, or check links extracted from a suspicious message without sending the message body to Guardio. Connect your Guardio account to review protection setup, scan activity, recent data leaks, linked-service counts, and security recommendations. You can also check whether an email address or phone number appears in known data breaches. Results describe known exposure; they do not reveal leaked passwords. When you ask, Guardio can guide setup of its protection in a supported browser and receive your feedback about these tools. Setup depends on browser support and available permissions. Account features depend on your access and plan. Guardio records each link check on your account, and a dangerous result appears in your account activity. A link result reports known threats at the time of the check. It is not a guarantee that a site or an entire message is safe. The plugin does not scan local executable files or provide continuous browser protection merely by being connected.
Atbash AI
Connect an Atbash agent, manage its name, purpose, risk, or active state with browser approval, and check explicitly submitted actions against Atbash safety policies. The desktop helper keeps agent private keys in local configuration and uses them locally for signing. Safety policy enforcement applies to the skill's decision to proceed with an explicitly checked action: HOLD, BLOCK, and errors stop that workflow. This package does not automatically intercept other tool calls or execute submitted commands. Requires Node.js 22.13.0 or newer on macOS arm64, Linux x64/arm64 (glibc), or Windows x64.
Breeze Security
Breeze helps authenticated users explore their organization's security posture, connected integrations, scan status, tenants, and entity risk data through ChatGPT and Codex.
ailuntz
Keybook 是供本机已授权任务使用的凭据目录。首次使用时，告诉它你现有凭据 YAML 文件的绝对路径；它会在本地检查格式，只保存路径。以后无需反复粘贴密码或 API Key，直接让 AI 继续原任务即可。 需要登录网站时，可查找 web 分类中的用户名、邮箱、指定密码变体和 PIN；需要本机授权时，可分别查找 system 中的 macOS 登录密码或登录钥匙串密码。发布 npm 包、上传 Hugging Face 模型或数据集、推送 Docker 镜像、操作 GitHub/GitLab、部署 Cloudflare 时，可按名称查找 services 中对应的令牌。调用 OpenRouter、ModelScope 等模型 API 时，查找 llm 中的密钥；连接已授权的 SSH 主机时，查找 ssh 中的私钥文件路径。实际可用条目取决于你自己的 YAML，Keybook 不会假定某个服务已配置。 Keybook 先搜索条目名称，再用随包脚本把选定值注入可信目标程序的环境变量，并继续任务。list/search 只返回名称，get 始终遮盖值，run 隐藏子进程输出并只报告状态；不会在对话中展示明文。SSH 条目是文件路径，由目标程序传给 SSH 客户端。你自行维护本地 YAML；插件不会创建、加密或同步它，也不提供浏览器自动填充。需要本机终端与 Node.js 22+。
Leakdata.io
Understand a data breach and choose your next steps. Look up a company or website to see known breaches, the types of data exposed, and practical first steps. Browse recent breaches in LeakData's public catalog without connecting an account. Connect your LeakData account to check its verified email address, review existing monitoring results, and work through the most urgent steps first. Ask ChatGPT to guide you through your action plan. When you confirm you finished a step, save it as done; reopen it on request. This records progress in LeakData; it does not change passwords or settings on other services. For your company, learn what domain monitoring checks and how to verify ownership. This chat feature can show public catalog entries for a supplied domain, but does not search staff email addresses or turn on monitoring. Results cover the data LeakData holds. No match does not mean that an account or company is safe. Do not share passwords in chat; the plugin does not show raw leaked records or search another person's private breach data. You can ask questions in your preferred language. Cards follow ChatGPT's interface language when supported, with English as the fallback.
Above Security
Above brings your identity-security data into ChatGPT. Ask in plain language to list open incidents and high-risk identities; inspect incident timelines, evidence, and insights; look up devices, monitored applications, and audit logs; and run investigative searches across your tenant. You can also triage from chat — set incident status, verdict, and resolution notes, or add an identity to the watch list. Access requires an Above Security account and is scoped to your role and company; read and write are governed by explicit OAuth consent.
Skarn Software OÜ
Skarn is a command-line security scanner for AI coding sessions. The audit skill, skarn-audit, teaches the model to run the skarn CLI on this machine and act on what it reports. The skill runs two passes. skarn assess reads the session transcripts that AI coding assistants have already written to disk and reports leaked credentials, prompt injection, exfiltration attempts, and the chains that connect them; each finding carries the standards mappings that apply to it, drawn from MITRE ATLAS, the OWASP Top 10 for LLM Applications 2025, the OWASP Agentic Top 10, and CWE. skarn vet reads the assistant configuration files on the same machine and reports risky declarations: hooks, MCP servers, and permission grants that could run remote code or send data off the machine. Both passes run entirely on this machine and make no network call. Neither modifies a session transcript or a configuration file, and every credential value in the output is masked, so a finding names a rule and a masked preview and never the credential itself. Skarn surfaces findings and does not rotate a credential or edit a configuration on its own: the skill proposes, and the user decides. The one thing the skill can write is a baseline entry recording a finding the user has explicitly confirmed as a false positive. The skill runs a local shell command, so it is offered in Codex. It does not carry the skarn binary, which is installed separately: brew install skarn-security/tap/skarnnpm install -g @skarn-security/skarnRelease binaries: https://github.com/skarn-security/skarn-dist/releases/latest When the binary is not installed, the skill says how to install it and stops rather than inventing a result. skarn assess and skarn vet need no account and no license. Skarn also ships a hook that inspects a tool call before Codex runs it, and a skill that reconstructs recent work from session history. Both are distributed through the plugin marketplace at https://github.com/skarn-security/agent-guard. Documentation: https://getskarn.com/integrations/Support: hello@getskarn.com
AttackerView
Ask whether a website is set up securely and get an answer in seconds. AttackerView checks a domain's security headers and Content Security Policy, clickjacking protection, the HTTP to HTTPS redirect, TLS configuration, DNS, DNSSEC and CAA records, and email authentication (SPF, DKIM, DMARC, MTA-STS). Each problem comes with its severity, the evidence observed and how to fix it. For developers, founders and IT teams who want a quick read on a site's configuration, their own or one they depend on. The check is passive: one request to the homepage plus public DNS and TLS lookups. It does not log in, crawl or send attacks, so it cannot find vulnerabilities in the application itself. No account is needed.
Samautomation
Check a suspicious email, text message, WhatsApp message, link or screenshot for scam signals. You get your own message back with every red flag quoted and explained, a verdict (almost certainly fake, something is off, no red flags found, or unable to judge) and the next steps. It never says a message is safe and never opens the suspicious link. Built for the Netherlands and Belgium; explanations are in Dutch. No account needed, and messages are not stored.
Greenlit Books
Ask whether an AI agent, MCP server, coding assistant or AI product is safe to use, and get a straight verdict from Greenlit Books, with the setup steps that make it safer and a link to the full review. Each verdict comes from reading the product's own code, docs and settings, and says which version it read and when. There are more than 800 of them, covering coding assistants, automation tools and hundreds of MCP servers. It also checks statements about AI agents against a ledger of 564 claims, each sorted by what it rests on (a published result, the book's argument, a method you can run, or the author's own account) and each with a citation. Where a claim rests on an outside source, the result says what that source does and does not establish. Ask it how or why about working with AI agents and it answers from Greenlit's books: it finds the passage in the free first chapters, the concept pages, the glossary and the claim ledger that bears on your question, quotes it with a link to the exact section, and points you to the page where the rest of the topic is. And when you want to go deeper, it finds the right practical book on working with AI, with a free first chapter to read. Read-only. No account and no sign-in. Nothing you ask is stored beyond short-lived request logs.
Drata Inc.
Use natural language to work with your Drata governance, risk, and compliance data. Search and list controls, policies, evidence, risks, vendors, personnel, devices, and monitoring tests. Create and update supported records where your Drata permissions and granted OAuth scopes allow. Connect your Drata account with OAuth. Access is limited by your Drata role and granted scopes.
