SnailSploit/Claude-Red

The problem

General chat assistants can discuss SQL injection or ADCS in the abstract, but they have no structure for an actual engagement: no ordered workflow, no tooling notes, no escalation paths. The result is generic advice you still have to reshape yourself. claude-red packages 78 offensive-security skills as drop-in SKILL.md files, each scoped to one attack surface, so that mentioning a target technique loads the corresponding methodology into context.

What's different

Compared with pasting a long prompt or keeping your own notes, the interesting design choice is packaging knowledge as per-topic skill files that load on demand, so unrelated skills do not consume context. Compared with large security knowledge bases, it is narrow and operational rather than encyclopedic. The breadth of the index is unusual: 23 categories spanning web, wireless, AD, cloud, exploit development, container escape, and AI security. The README itself is mostly an index of filenames and one-line descriptions; it describes methodology in prose and does not present measured effectiveness, so how deep each skill actually goes cannot be judged from the repository page.

Who it's for

Authorized red teamers, bug bounty hunters, CTF players, and security researchers already using Claude who want a starting methodology rather than blank-page prompting. It is most useful to people who can independently verify what the model says. It is not for people looking for a turnkey exploitation tool: these are text skills, not scanners or exploit code. Anyone without written authorization to test a target should look elsewhere; the repository frames itself around authorized engagements, and copy-pasted methodology does not change the legal picture.

Cost of adoption

No runtime beyond a Claude environment that supports skills. Installation is a git clone into ~/.claude/skills/claude-red, with a sparse-checkout option for single categories and an install.sh helper. MIT license, a Python install script, and no model weights or services. Maturity is the main risk: the project is roughly 234 days old, the README is thin on per-skill depth and maintenance process, and the star count is high relative to its age, which suggests attention arrived faster than review. Expect to read and vet individual SKILL.md files before trusting them on a real engagement.

Verdict

The project is young rather than established: about 234 days old, launched in early 2026, with 8,820 stars after a single trending appearance where it gained 770 stars in a day. That pattern is a spike, not a long-term track record, and there is no Hacker News discussion to corroborate the interest. The structure is sensible and the category coverage is broad, but the value lives entirely in the quality of individual skill files, which the README does not let you assess. Treat it as a well-organized prompt library to audit and adapt, not as an authority on offensive tradecraft.