Turn Claude into a specialist for specific attack surfaces
For complex security tasks like SQL injection or shellcode writing, Claude moves beyond listing information to follow the methodology of a domain expert. This repository collects SKILL.md files ready for the Claude skill system, which load automatically when a specific topic arises in conversation to manage context efficiently. This allows for more systematic exploration of attack paths during authorized red team operations or bug bounty analysis.
Skill structure that activates automatically based on conversation flow
Each skill is a structured file containing expert-level methodology for a specific attack surface. If a user mentions SQL injection in the conversation, the offensive-sqli skill loads automatically, while it does not burden the context when other topics are discussed. This on-demand loading approach reduces unnecessary token consumption while providing precise domain knowledge when needed. Skills cover various areas from web applications to wireless networks, cloud, and mobile, allowing for a switch to a situation-appropriate expert mode.
Attack surface coverage across 23 categories
The current library spans 23 categories: web applications, authentication and identity, Active Directory, wireless, cloud, mobile, IoT, infrastructure and red team, exploit development, fuzzing, reconnaissance, API security, containers and Kubernetes, CI/CD, cryptography, privilege escalation, post-exploitation, forensics and C2, supply chain, social engineering, network attacks, AI security, and utilities. Each category contains individual skills addressing specific techniques, tools, edge cases, and escalation paths for that area. For example, the wireless category covers attack methods for various protocols such as WPA2/3, EAP, BLE, Zigbee, Z-Wave, and LoRa, while the infrastructure category includes advanced red team operation techniques like EDR evasion and shellcode writing.
Installation and usage methods for different Claude environments
The most recommended method is to clone the repository to the ~/.claude/skills/claude-red path using the git clone command. This allows Claude to automatically recognize skills based on conversation triggers. If only specific categories are needed, git sparse-checkout can be used to selectively fetch only those folders. When using Claude Code, the content of SKILL.md can be passed to system files using the cat command, or in Claude.ai, it can be used by pasting the content into the project system prompt. Additionally, the install.sh script allows for interactive installation or installation by specifying particular categories.
Expansion plans and contribution guidelines
The library is currently undergoing multi-stage expansion with a target of approximately 130 skills. There is a roadmap of 9 stages including wireless, IoT, web basics and advanced, documentation and tool improvements, new category additions, and in-depth rewrites, with some completed and others in the planning stage. Contributions are welcome, and skills focused on a single surface are preferred over extensive overviews. CONTRIBUTING.md specifies the skill template, frontmatter standards, and review process, which can be referenced when adding new skills.
Legal and technical scope to verify before use
This repository is distributed under the MIT license and can be used freely, with attribution recommended. However, since these skills are for offensive security, i.e., aggressive security tasks, they must only be used for legal purposes such as authorized red team work, bug bounties, security research, CTF preparation, operator training, and systematic attack surface exploration. While skills guide methodology and tools, legal responsibility for actual attack execution lies with the user. Additionally, the accuracy and currency of skills may vary with changes in the threat environment, so it is advisable to check the latest trends in relevant technologies before important tasks.
