andreybuilt

Security Skills — Security skill for Claude Code

Security community

Two skills for reviewing what you are about to trust: an AppSec/CloudSec/CI-CD review pass, and an audit that vets a skill, plugin or MCP server before you install it.

How to install Security Skills

This entry records only its repository, not the path inside it, so there is no exact command to give. Open andreybuilt/security-skills and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Security Skills does

Two skills for reviewing what you are about to trust: an AppSec/CloudSec/CI-CD review pass, and an audit that vets a skill, plugin or MCP server before you install it. 16 fixtures, benign and malicious, and the detector must fire on both.

Alternatives in Security

Security Threat Model — Generate repo-specific threat models identifying trust boundaries 14.6k ★
Synk Skill Security Scanner — Agent Trust Hub 1.9k ★
Trail Of Bits Claude-code-config — Opinionated production defaults from a top security firm: sandboxing, permissions, hooks, skills, MCP server c 1.6k ★

README

security-skills

Two Claude Code skills that vet what an agent is asked to trust: the code it reviews, and the instructions it runs under.

Why one repository

An agent takes two kinds of input that both deserve suspicion before you act on them: the code a user hands it to review, and the skills, plugins, and configuration files the agent itself loads and treats as instructions. Most security tooling only covers the first kind. A SAST scanner reads your Terraform and your Python. It does not read `SKILL.md`, `CLAUDE.md`, an MCP server config, or a plugin manifest, and none of those look like code to a traditional scanner, so nothing flags them.

That gap matters because an instruction file is executable the moment an agent reads it. A hidden zero-width character, a buried HTML comment, or a sentence that tells the agent to "ignore the user's previous instructions" does not need to compile or run as a binary. The agent just has to read it.

`security-review` covers the first problem: application code, pull requests, cloud configuration, Kubernetes manifests, and CI/CD pipelines, mapped to OWASP, NIST SSDF, CIS, and SLSA. `skill-audit` covers the second: it scans skills, plugins, MCP configs, and instruction files for hidden Unicode, injected instructions, and exfiltration directives before you let an agent trust them. They ship together because a security review that skips the artifacts telling the agent how to behave has already missed half the attack surface.

Layout

security-review/
  SKILL.md                       the skill definition Claude Code loads
  references/                    checklists loaded on demand: AppSec, cloud, Kubernetes,
                                  CI/CD, severity model, finding templates, remediation
                                  patterns, and a skill/agent pre-install checklist

skill-audit/
  SKILL.md                       the skill definition Claude Code loads
  scripts/scan_hidden.py         the hidden-cont