ARTEX skills:a skills folder inside, skill repos outside.
An ARTEX skill is a package of instructions and files that an AI agent loads when it needs them. ARTEX workers read skills from a folder next to the program, and several outside repos package ARTEX-style methods as a skill for Claude Code and other agents. This page lists what each one offers, with links to the fork pages.
- GitHub stars on pojun-skill, 2026-11-16
- 21
- artex-skill renamed pojun-skill
- 2026-11-06
- API routes in the artexctl catalog
- 266
- skills in the agentic-soc defender pack
- 4
Skills inside ARTEX:a folder next to the program.
The ARTEX README, copied into most forks, says a release zip holds the artex program, start.sh, a skills folder and config.example.json. The Docker setup keeps ./skills and ./data on bind mounts, and an upgrade overwrites the program and the skills folder while your config and data stay. The ScopeWeaver README adds that a skill_dir setting points at the folder and that the ARTEX_SKILL_DIR variable overrides it.
Workers get skills next to the shell, HTTP requests, file reads, web search and remote MCP servers. The v0.3.22 release of 2026-10-19 fixed a skills folder that would not expand on Windows, in pull request 18 by adeljck. The artex-mcp README shows a web route, /api/skills/upload, that takes a skill as an uploaded ZIP file. BaaRai describes the feature as MCP servers, skills, memory and a workspace: reusable prompt and playbook material for the agent.
Skill repos for other agents:ARTEX methods as prompt files.
These repos are not forks of the ARTEX code. Each one packages a method or an admin guide so that Claude Code, ZCode, Cursor or another agent can follow it.
chu0119/pojun-skill
AGPL-3.0, last push 2026-11-06
Formerly artex-skill. It was renamed on 2026-11-06 when the project became Pojun and the skill moved to v2.0. It is one SKILL.md plus reference files, with no infrastructure and no dependencies. The README lists a planner, worker and auditor rotation, three markdown ledgers for assets, facts and findings, evidence levels L1 to L4, and four safety layers that start with a written scope. It says it is for authorized testing only. It works as an Agent Skill in Claude Code and ZCode, and as a rules file or a pasted prompt in other agents.
ADENH/artex-skill
AGPL-3.0, last push 2026-10-10
A separate repo that carries the earlier, ARTEX-era version of the same README: the three-role protocol, the three ledgers and the four safety layers, under the artex-skill name.
EnglandLobster/artex-cli
AGPL-3.0, last push 2026-10-30
artexctl, a command line for the ARTEX web API, plus an agent skill in cli/skill/artex that documents the commands for Claude Code, ZCode and similar agents. It is a Python 3.10 standard-library tool with a catalog of 266 backend routes, can reach a server through an SSH tunnel and saves its token with owner-only file permissions. The skill loads one topic at a time: tasks, sessions, findings, assets, approvals or platform admin.
artex131071-andrey/artex-skills
last push 2026-11-01
A name match, not the pentest tool. Its README, in Russian, describes shared rules and 11 working skills for Codex and Claude on Windows, and does not mention the ARTEX framework.
Skills shipped inside forks:bundled with the platform.
Some forks add their own skills to the folder. These are the ones whose README says so.
wpsec/ARTEX-Pro
last push 2026-11-16
The README describes the ARTEX-Pro iteration of the last source. It lists built-in skills for web, API, code review and internal-network work, where web and API testing calls a bundled OWASP ZAP over MCP.
TATHPE/ARTEX-Endfield
last push 2026-10-18
Adds the REA reverse-engineering MCP server and ships a copy of the reverse-engineer-anything skill, MIT licensed, which is only visible when the rea MCP is enabled.
azrielnan/artex-ko-main
last push 2026-10-30
A Korean-oriented build. It bundles a rea-reverse-engineering skill and an optional REA MCP for analyzing local programs, packages and source trees.
chu0119/Pojun
offensive-class
The platform behind pojun-skill. The pojun-skill README says the platform ships 21 scenario skills and a 48-article SRC knowledge base. This guide links only its fork page.
Fork page on artex.shFork page only: offensive-class builds are not linked
Skills for defenders:reading ARTEX activity, not running it.
Two repos use the skill format on the defense or teaching side.
Ed-Marcavage/agentic-soc
MIT, last push 2026-11-15
Four skills for the coding agent you already use. Three review your code for the weaknesses that autonomous pentest harnesses such as ARTEX use in real intrusions: exposed-attack-map-review, account-flow-abuse-review and cloud-credential-reach-review. The fourth, agentic-intrusion-hunt, reads CloudTrail and web access logs and says whether an attacker agent is in them. The README says it is current as of 2026-11-15 and built from nine reports published between 2026-10-21 and 2026-11-13. It installs as a Claude Code plugin or with the Skills CLI.
0x7377/agentic-ai-offensive-security-lab
Apache-2.0, last push 2026-11-15
A Korean-language class kit that combines OpenCode, an ARTEX class environment, a CVE MCP server and a class Skill ZIP. It says it is for local lab targets only and leaves out the slides and instructor answers.
Keep reading:related pages and sources.
Related pages
Follow the ARTEX story.
New forks, releases and reports, dated and sourced.
