███████╗███████╗ ██████╗██╗   ██╗██████╗ ██╗████████╗██╗   ██╗
██╔════╝██╔════╝██╔════╝██║   ██║██╔══██╗██║╚══██╔══╝╚██╗ ██╔╝
███████╗█████╗  ██║     ██║   ██║██████╔╝██║   ██║    ╚████╔╝
╚════██║██╔══╝  ██║     ██║   ██║██╔══██╗██║   ██║     ╚██╔╝
███████║███████╗╚██████╗╚██████╔╝██║  ██║██║   ██║      ██║
╚══════╝╚══════╝ ╚═════╝ ╚═════╝ ╚═╝  ╚═╝╚═╝   ╚═╝      ╚═╝

███████╗██████╗ ███████╗ ██████╗██╗ █████╗ ██╗     ██╗███████╗████████╗
██╔════╝██╔══██╗██╔════╝██╔════╝██║██╔══██╗██║     ██║██╔════╝╚══██╔══╝
███████╗██████╔╝█████╗  ██║     ██║███████║██║     ██║███████╗   ██║
╚════██║██╔═══╝ ██╔══╝  ██║     ██║██╔══██║██║     ██║╚════██║   ██║
███████║██║     ███████╗╚██████╗██║██║  ██║███████╗██║███████║   ██║
╚══════╝╚═╝     ╚══════╝ ╚═════╝╚═╝╚═╝  ╚═╝╚══════╝╚═╝╚══════╝   ╚═╝

⚡ Security Specialist

Skills that reason about a codebase for vulnerabilities and set up automated security scanning in CI, for any agent.

Validate skills License: MIT Agent skills Format Tooling

Table of contents:

Skills
How they fit together
Install
Usage
Optional agent profiles
Validate
Layout
Status and limits

Important

Principle: never auto-apply a security fix or scanning change, and never invent a finding, a credential, or a CVE that was not actually observed.

These skills scan a codebase for vulnerabilities the way a security researcher would, and set up automated dependency, static and dynamic scanning in CI/CD. They are grounded in the OWASP Top 10 categories and common framework-specific anti-patterns, and follow the open Agent Skills format. They work in Claude Code, Codex, Cursor, Copilot, OpenCode, or any SKILL.md-compatible agent.

🧩 Skills

Skill	Purpose
`security-ci-scanning`	Sets up automated security scanning in a CI/CD pipeline: dependency (SCA) audits, static (SAST) scanning, and dynamic (DAST) scanning of a running API or web app.
`security-code-review`	Reasons about a codebase for security vulnerabilities the way a security researcher would: tracing data flow, auditing dependencies and secrets, and reporting findings with severity and confidence.
`security-specialist`	Router and shared guardrails for application security work: reasoning about a codebase for vulnerabilities and setting up automated security scanning in CI/CD.

🗺️ How they fit together

flowchart TD
    R["⚡ security-specialist<br/>routes work and enforces shared guardrails"]
    R --> A["🔍 security-code-review<br/>vulnerability scan, severity and confidence"]
    R --> B["🛠️ security-ci-scanning<br/>dependency, static and dynamic scanning in CI"]

Most requests start at the router. A manual review usually comes first; automating it in CI comes once the team is happy with what it finds.

📦 Install

Tip

Install the router skill plus every sibling skill it links to; routers reference siblings by relative path.

Claude Code plugin (installs the skills and any agent profiles together):

/plugin marketplace add KinwiseNZ-DevOps/security-specialist
/plugin install security-specialist@security-specialist-marketplace

ChatGPT desktop plugin (local, skills and bundled hooks):

This repo has a portable plugin.json manifest and a repo marketplace at .agents/plugins/marketplace.json. Open this repo as a trusted project and restart ChatGPT desktop, then install Security Specialist from the Plugins Directory. Review and trust bundled hook scripts before they run.

Or copy or symlink the skills/ subfolders you want into your agent's skills directory:

Agent	Where skills live
Claude Code	`~/.claude/skills/` or a project's `.claude/skills/`
Codex · Cursor · Copilot · OpenCode · Windsurf	the agent's own SKILL.md discovery path

To install supplied Claude Code, GitHub Copilot or Codex profiles into a project, run python scripts/install_agents.py --host claude-code --target PATH. Each repo supplies one main workflow agent and may supply one helper per host. Other hosts receive the skills only. If no profile files are supplied for that host, it makes no changes.

Always on

In Claude Code, Security Specialist loads a short ruleset into every session as soon as the plugin is installed (never auto-apply a fix, rate severity and confidence separately, never guess a secret). It tells you its state at the start of each session and costs roughly 450 tokens.

Mode	What it does
`on` (default)	The rules apply in every session
`auto`	The rules apply only in projects that look relevant: a `package.json` or `requirements.txt` is found
`off`	The rules do not load

Say "stop security mode" to pause it for the current session, and "start security mode" to resume.
Ask Claude to set the saved mode to on, off or auto. It writes that word into ~/.claude/.security-specialist.mode. You can also edit that file yourself, and deleting it returns to the default (on).
It needs Node on the machine, because the start-up hook runs a small Node script. Without Node the hook quietly does nothing.
Other agents have no start-up hook here. Paste skills/security-specialist/references/always-on-rules.md into that agent's instruction file instead.

Safety checks

The same hooks run a few checks while you work. Each one only asks, warns or adds a reminder; none changes your files on its own.

Check	What it does	On by default
`dangerGuard`	Asks before risky commands: force pushes, `git reset --hard`, `rm -r`, dropping data, `docker compose down -v`, publishing a package	Yes
`secretsGuard`	Asks before writing keys, tokens or `.env` files	Yes
`packageGuard`	Asks if a package about to be installed does not exist on PyPI, npm or pub.dev, because the name may be invented	Yes
`licenseGuard`	Warns after installing an npm or PyPI package with a GPL-family licence	Yes
`editSecurity`	Reminds Claude about risky code it just wrote: `eval`, SQL built from strings, unescaped HTML, tokens in SharedPreferences	Yes
`promptGuard`	Warns when pasted text looks like it is trying to steer the agent	Yes
`verifyReminder`	Tells you at the end of a turn if code changed but no tests, lint or analysis ran	Yes
`linkGuard`	Warns about links that return "not found" in Markdown, HTML and JSX files written	No
`styleGuard`	Warns when a reply is over 200 words or uses an em or en dash (built for AI Kiwiniser's style)	No
`sessionLog`	Writes tool names and file names, never content, to `~/.claude/logs/security-specialist/`	No
`autoCommit`	Saves tracked changes as a local commit at the end of a turn. It never pushes	No

Switch any check with ~/.claude/.security-specialist.features.json, for example {"packageGuard": false, "sessionLog": true}, or ask Claude to write it. The saved mode (on, off or auto) applies to every check. If several of these plugins are installed, each risky action is checked once, not once per plugin.

Privacy: packageGuard and licenseGuard send package names, and linkGuard sends the links it checks, to the public registry or site. Nothing else leaves your machine. Turn those checks off if that is a concern.

🚀 Usage

Tip

You never need to memorise skill names. Just describe the problem in plain words; the router picks the right skill automatically. But if you want short, reliable phrases that always work, use these:

The flow (this is all most people need):

1. "Is this code secure?"                       → Scans and reports findings with severity and confidence
2. "Set up automated security scanning"          → Proposes CI scanning as a pull request
3. "Fix this SQL injection"                      → Reviews the specific code and proposes a patch to approve

Short phrases that trigger each skill:

Just say…	What you get	Skill used
`Scan this repo for vulnerabilities`	A findings report with severity and confidence, grouped by category	security-code-review
`Check for exposed secrets or API keys`	A secrets and exposure scan, with safe placeholders correctly ignored	security-code-review
`Audit our dependencies for known CVEs`	A dependency audit against the vulnerable-package watchlist	security-code-review
`Add Dependabot/CodeQL/DAST scanning to our CI`	A pull request with the scanning workflow and a list of required secrets	security-ci-scanning

Note

Skills also trigger on their own when your request matches. In Claude Code you can also invoke one explicitly with /<skill-name>, but the plain phrases above are shorter and work across every agent.

Optional agent profiles

This section applies only when agent definitions were supplied during setup. No agent profiles are supplied yet; all three skills run directly. See AGENTS.md for the dated note.

✅ Validate

python scripts/skills_tool.py check

Checks: frontmatter, name/folder match, description length and trigger, body under 500 lines, every relative link resolves (including README.md and AGENTS.md), agent profile naming with no more than two profiles per host and exactly one -workflow main agent, agent skills: entries that exist and no skills/<name>/SKILL.md paths in agent bodies, plugin manifests (.claude-plugin/plugin.json and marketplace.json: required fields, licence matches LICENSE, agents lists every Claude Code profile, no top-level agents/ folder), at least 3 eval cases per skill, catalogue freshness, no project-specific terms from reference-content/strip-terms.txt, and no em or en dashes in file names or in any text outside reference-content/. After the starter skill is deleted it also rejects leftover placeholders and setup-only paths that are tracked by Git or not ignored. Stdlib only, no dependencies.

🗂️ Layout

Repository layout

skills/<name>/SKILL.md      one folder per skill (+ references/ where needed)
.claude-plugin/             plugin.json and marketplace.json (Claude Code plugin)
hooks/                      always-on rules and safety checks for Claude Code
evals/cases.jsonl           3+ behavioural cases per skill
resources/agents/             optional host profiles copied by installer
scripts/install_agents.py     copy optional agent profiles into a project
scripts/skills_tool.py      validate, regenerate catalogue (stdlib only)
.github/workflows/          CI runs `skills_tool.py check` (free on public repos)
AGENTS.md                   agent-facing rules for this repo
SETUP.md                    local checklist; ignored and kept off GitHub
reference-content/          private source material; ignored and kept off GitHub
templates/                  local skill scaffolding; ignored and kept off GitHub

📌 Status and limits

Note

20 October 2026: skills were authored from condensed, generalised security references (OWASP-style anti-patterns, a code-reasoning review workflow, and CI scanning setup guidance); none has been run against a real vulnerability yet. Reason: first draft awaiting real-world testing.

The vulnerable-package watchlist in security-code-review is a dated snapshot; check current advisories for anything it does not cover.

Licensed under MIT (see LICENSE).

⚡ Built on the Agent Skills open format · grounded in the OWASP Top 10