Today's drop: 1 new tool, 3 listing updates
The directory lists 150 hand-checked tools. New today: Claude-BugHunter, a free, open-source skill bundle that teaches Claude Code and other coding agents how bug bounty hunters and red teamers work, from recon to a report that survives triage. Updated: ChatGPT and OpenAI Codex (GPT-6.2 Astra and Sol, plus Codex CLI 0.170.0) and Claude Code (v2.1.327). AI for IT & Cybersecurity grows to 66 tools; we retired one listing to stay at 150. Pricing only where the official page states it.
Previous drop: Nov 17, 2026 →
Get the weekly dropThe best new AI tools, one email a week. Free.
New in the directory 1
Browse all 150Tool
Category
Price
Status
Link
Claude BugHunterFree, open-source (MIT) bundle of 83 agent skills and 15 slash commands for bug bounty hunting and external red-team work, built by security researcher Sachin Sharma. It packs a five-phase bug bounty methodology, 58 hunt skills for web bug classes such as IDOR, SSRF, OAuth, JWT, file upload and GraphQL (drawn from 681 disclosed HackerOne report patterns, 433 of them individually cited), recon and OSINT skills, attack chains for internet-facing enterprise systems (Microsoft 365 and Entra ID, Okta, SharePoint, vCenter, SSL VPN appliances) and reporting skills with a seven-question triage gate that checks scope and accepted impact before you submit. Skills load automatically when you describe what you are testing in plain English. Install it as a Claude Code plugin (/plugin marketplace add elementalsouls/Claude-BugHunter) or with the install script, which also copies the skills to OpenCode, OpenAI Codex CLI, Hermes Agent and Google Antigravity; the slash commands and hunt engine are Claude Code only. Optional Burp MCP wiring and, since Nov 18, 2026, a HackerOne MCP server. Intended only for assets you own or are authorized to test (bug bounty scope, pentest engagements, CTFs, your own lab); the author notes Anthropic's safeguards may refuse some offensive requests unless you join Anthropic's free Cyber Verification Program. Public since May 2026; about 10,300 GitHub stars, with about 2,300 added in one day on Nov 18, 2026 when it topped GitHub Trending.
Price Free and open source (MIT); needs your own Claude Code or other agent subscription
Why this one
- Claude-BugHunter was the most-starred new repository on GitHub Trending this morning (about 2,300 stars in a day, about 10,300 in total). It is a set of 83 agent skills and 15 slash commands by security researcher Sachin Sharma: a bug bounty methodology, 58 hunt skills for bug classes such as IDOR, SSRF, OAuth and JWT built from 681 disclosed HackerOne report patterns, recon and OSINT, and a triage gate that asks whether a target is in scope before you report. You describe what you are testing in plain English and the matching skill loads. It installs as a Claude Code plugin, and the same skill files work in OpenCode, Codex CLI, Hermes Agent and Google Antigravity. A HackerOne MCP server was added on Nov 18, next to the existing Burp MCP hookup. For anyone breaking into security, the honest use is practice: run it against DVWA, OWASP Juice Shop or Hacker101 in your own lab and read why each skill picks the next step. Only point it at assets you own or are authorized to test, and expect Claude to refuse some offensive requests unless you are in Anthropic's Cyber Verification Program.
Listing updates
Checked on Nov 18, 2026 against official pages and release notes.
- ChatGPT and OpenAI Codex: GPT-6.2 Astra (OpenAI's new frontier model) and GPT-6.2 Sol (its everyday model) arrived Nov 17 for Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex; Enterprise and Edu admins turn it on. In the API,
gpt-6.2-astracosts $8 per million input tokens ($0.40 cached) and $40 per million output tokens, 20% below GPT-6 Astra, and Sol keeps GPT-6.1 Sol's prices. OpenAI notes both models were sometimes more likely to leave out obstacles that blocked their work, and says it added instructions in Codex to reduce that. - Codex CLI: 0.170.0 (Nov 18) adds realtime voice calls in the terminal UI, fixes MCP OAuth credential races during login and logout and keeps existing sandbox restrictions when permissions change or a child thread starts.
- Claude Code: v2.1.327 (Nov 17) asks before
/install-github-appchanges a repository, addsclaude mcp add-json <name> -so server tokens can come from stdin instead of the command line, lists broken skill and agent files, ignored permission settings and hook matchers that never match in/status, and fixesclaude mcp addandremovedeleting other keys from.mcp.jsonand/exportoverwriting a file without asking.
Category changes
- AI for IT & Cybersecurity grows from 65 to 66 tools with Claude-BugHunter.
- Image drops from 7 to 6 tools: to stay at the 150-tool cap we retired Krea. Leonardo.Ai, Midjourney and Ideogram cover AI image generation for most of our readers, and our Krea listing never went beyond a one-line description. Its old page redirects to the category.
Checked, not listed
- SkillEvaluator (NVIDIA): still trending (about 3,450 stars), still a quality framework for skill authors; SkillSpector covers skill security.
- MAI Code 2 and Code 2 Flash (Microsoft AI): on the Hacker News front page this morning; we have not checked them yet.
- Routine: Qwen Code 0.28.1 is a preview only; n8n 2.48.5 patch release; Wazuh 4.14.11 (rc2) and Ollama v0.40.17 are still release candidates.