SnailSploit/Claude-Red

What it is

A GitHub repo (SnailSploit/claude-red, MIT-licensed, by Kai Aizen) of SKILL.md files for Anthropic's Claude Skills system, each priming Claude as a domain specialist for one offensive-security surface. The summary covers the concept and use cases; what it omits is the concrete install/use mechanics and the scope target.

What to do

Three install paths, verbatim from the page.

Full library into the Claude skills directory (auto-loads on conversational trigger — e.g. mentioning SQL injection loads offensive-sqli):

git clone https://github.com/SnailSploit/claude-red ~/.claude/skills/claude-red

Single category via sparse-checkout, then feed a skill as a system prompt:

git clone --filter=blob:none --sparse https://github.com/SnailSploit/claude-red
cd claude-red && git sparse-checkout set Skills/web Skills/active-directory
cat Skills/web/offensive-sqli/SKILL.md | claude --system-file -
cat Skills/active-directory/**/SKILL.md | claude --system-file -

Interactive installer, with flags for explicit target or a single category:

./install.sh                           # interactive
./install.sh --target ~/.claude/skills # explicit target
./install.sh --category web            # single category

Or paste a SKILL.md directly into a Project's system prompt.

Scope / state

Stated target is ~130 skills across 23+ categories — the repo is being built out in phases (see CHANGELOG.md), so at any given time it is partial against that goal. Category directories present include: web, auth, active-directory, wireless, cloud, mobile, iot, infrastructure, exploit-dev, fuzzing, recon, api, container, cicd, crypto, privesc, post-exploitation, forensics, supply-chain, social-engineering, network, ai, utility. If you need a specific surface, check the directory exists before relying on it.

Caveats

Contributions favor "focused, single-surface skills over monolithic overviews" (CONTRIBUTING.md holds the frontmatter standard and review process), so depth is per-skill and uneven while phases roll out. Content is authored/community-reviewed methodology, not tested tooling — treat each skill as a checklist prompt, not a guarantee of correctness for a given engagement.