Guide · Agent skill
How to use the API Security Best Practices agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The API Security Best Practices skill is a set of instructions that teaches an AI coding agent to review API code against a security checklist: how callers are identified, whether they may act on a specific object, whether input is strictly validated, and how abuse such as brute force or rate abuse is handled. It is written for backend developers who are adding a protected endpoint or reviewing an existing route. After this guide you will know how to install the skill, how to ask for it by name in plain language, and what it will and will not do for you.
API Security Best Practicessickn33/agentic-awesome-skills · MIT Download the skill Quick start
- 1Download the ZIP. Download the skill ZIP from this page and unzip it. The skill folder contains SKILL.md.
- 2Place the folder. Move the unzipped folder into ~/.claude/skills/ for personal use or .claude/skills/ inside a project.
- 3Or upload in the apps. In the Claude apps, go to Settings → Capabilities → Skills and upload the skill.
- 4Describe a concrete task. Ask in plain language about a specific route, such as adding a protected endpoint or reviewing object access.
- 5Provide context. Share the caller and tenant model, identity provider, token contract, database and runtime details.
- 6Adapt and test the output. Adapt the code sketches to your installed library versions and write tests for cross-user and cross-tenant cases.
What agent skills are and what this one does
An agent skill is a folder containing a SKILL.md file with instructions that an AI agent loads when a task matches the skill's description. This skill's description is 'Implement secure API design patterns including authentication, authorization, input validation, rate limiting, and protection against common API vulnerabilities'.
When it is active, the agent looks at the request boundary from caller identity through authorization, validated input, storage and the response. It favours the identity provider and data model your application already uses rather than introducing a second authentication system.
- Authentication: prefer the established provider or session middleware; for an HMAC JWT contract, require a fixed algorithm, exact issuer and audience, and required claims.
- Authorization: apply the owner and tenant condition inside the database mutation to avoid check-then-write races, and do not add an implicit admin bypass.
- Input validation: reject partial numeric parses and unexpected properties, and use the validated value in handlers.
- Abuse control: distinguish per-user quotas, per-IP controls and upstream budgets, and do not claim DDoS protection.
- Common mistakes it warns about: reusing access tokens as refresh tokens, SSRF through outbound URLs, and unsafe file uploads.
Install the skill from the ZIP
Download the ZIP for this skill from this page and unzip it. The skill is a single file, SKILL.md. Put the unzipped skill folder in one of two places:
Personal install: ~/.claude/skills/ so the skill is available across your projects.
Project install: .claude/skills/ inside a specific project, so it applies only to that repository.
In the Claude apps you can instead upload the skill under Settings → Capabilities → Skills.
- Personal folder: ~/.claude/skills/
- Project folder: .claude/skills/ inside your project
- Claude apps: Settings → Capabilities → Skills, then upload
Trigger the skill with natural requests
You do not need a special command. Describe the task in plain language, and mention API security, authorization, validation or the specific route. The skill is meant for work on a concrete endpoint or defect, so name the route and its callers where you can.
The skill's own guidance says that for a concrete defect it starts from the failing route and its callers, and that it should not deploy unrelated security infrastructure. Keep requests focused.
- 'Add a protected DELETE /posts/:id endpoint and make sure users can only delete their own posts.'
- 'Review this Express route for object-level access control problems.'
- 'Replace the loose request parsing in this profile update handler with strict validation.'
- 'Check how this JWT middleware verifies tokens and tell me what is missing.'
- 'This endpoint fetches a URL from user input. What SSRF risks should I handle?'
Example: a checklist review of an endpoint
Suppose you ask the agent to review a route that deletes a post by ID. With the skill active, the agent checks who the caller is, then whether the query scopes deletion to that caller and tenant. Its guidance includes a Prisma-style sketch that deletes with the owner and tenant predicate in the same statement, rather than looking up the post first and deleting it afterwards.
The sketch is illustrative. The ID and tenant types must match your own schema, and you should test it against the library versions you have installed.
async function deleteOwnedPost(prisma, postId, principal) {
const result = await prisma.post.deleteMany({
where: { id: postId, userId: principal.subject, tenantId: principal.tenantId }
});
return result.count === 1;
}What you need before using it
The skill itself requires no API keys or accounts. The code examples assume a Node.js runtime and reference Express, jsonwebtoken and Zod. The ownership sketch assumes Prisma or another database layer.
The skill's inputs list asks for the routes, caller and tenant model, identity provider, token contract, runtime, locked dependency versions, database schema, proxy topology and authorized test scope. Having these ready makes the review more accurate.
- Node.js runtime for the example code
- Express, jsonwebtoken and Zod, as referenced by the examples
- Prisma or another database layer for the ownership query
- Use synthetic test identities in a test environment
Safety notes: scripts and network
The skill runs no scripts and uses no network access. It is guidance for an agent, not a tool that scans or changes systems by itself. Any code changes, tests or scans are carried out by you or your agent in your own environment.
The skill states that production scans, account writes and message sends need their own separate authority. It provides no scanner, so do not expect it to probe a live service.
- Runs scripts: no
- Uses network: no
- Do not use it for production penetration testing or live scanning
- Test only against scope you are authorized to test
Tips for getting useful results
Give the agent the real token contract and data model. The guidance prefers your existing provider and schema over generic patterns, so a precise description produces more relevant advice.
Treat every code sketch as a starting point. Adapt it, then write tests for the cases the skill lists, such as a user accessing another user's object, the same ID in another tenant, and bulk endpoints.
- Name the route and the callers involved.
- Share the identity provider and token format you actually use.
- Ask for tests covering cross-user and cross-tenant access.
- Verify every API against the versions in your lockfile before integrating.
What people use it for
- Adding a protected endpoint that only lets users change their own records.
- Reviewing an existing Express route for object-level access control gaps in a multi-tenant app.
- Replacing loose request parsing with strict Zod schemas and validated values in handlers.
- Checking a JWT verification middleware for a fixed algorithm and issuer/audience checks.
- Reviewing an endpoint that fetches user-supplied URLs for SSRF exposure.
Frequently asked questions
How do I install the api-security-best-practices skill?
Download the ZIP from this page and unzip it. Place the folder in ~/.claude/skills/ for personal use or in .claude/skills/ inside a project, or upload it in the Claude apps under Settings → Capabilities → Skills.
Does the API security skill require an API key or account?
No. The skill itself requires no API keys or accounts. The code examples assume Node.js with Express, jsonwebtoken and Zod, and the ownership sketch assumes Prisma or another database layer.
Is the skill free to use?
The skill is published under the MIT licence. This page does not list pricing, so check the licence text in the source repository for the terms.
Does the skill scan my API for vulnerabilities?
No. It is guidance for an agent that reviews code against a security checklist. It runs no scripts, uses no network, and provides no scanner. The skill states that production scans need separate authority.
Can I use it for penetration testing?
No, not for production penetration testing or live scanning. The skill is aimed at code design and review, and it says production scans need their own authorization.
Why does the agent not just give me a finished secure service?
The examples are sketches for Express, jsonwebtoken and Zod that need adapting to your application and tested against the library versions you have installed. The skill names adapters such as the database transaction layer rather than bundling a full runnable service.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
This skill suits backend developers who want an agent to review a protected endpoint, object access rules or input validation against a consistent checklist. It is a poor fit for anyone who needs live scanning, penetration testing or a finished, runnable secure service, and for those cases a dedicated security tool or a specialist is the better choice.
Download the skill