AgentStack
SKILL verified MIT Self-run

Auth

skill-kid-sid-claude-spellbook-auth · by kid-sid

Use when implementing login flows, issuing or validating JWTs, setting up OAuth2/OIDC with a provider, designing role-based or attribute-based access control, securing API endpoints, or handling token refresh and revocation.

No reviews yet
0 installs
16 views
0.0% view→install

Install

$ agentstack add skill-kid-sid-claude-spellbook-auth

✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • Prompt-injection patterns
  • Secret / credential exfiltration
  • Dangerous shell & filesystem operations
  • Untrusted network calls
  • Known-malicious package signatures

What it can access

  • Network access Used
  • Filesystem access No
  • Shell / process execution No
  • Environment & secrets Used
  • Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

Are you the author of Auth? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

Authentication & Authorization

Patterns for identity, token management, and access control across web APIs and services.

When to Activate

  • Implementing login, logout, or registration flows
  • Issuing, validating, or refreshing JWTs
  • Integrating an OAuth2/OIDC provider (Google, GitHub, Auth0, Keycloak)
  • Designing role-based (RBAC) or attribute-based (ABAC) access control
  • Securing REST or GraphQL endpoints with middleware/guards
  • Handling token revocation, rotation, or blacklisting
  • Auditing an existing auth implementation for security gaps

Core Concepts

Authentication vs. Authorization

| Concept | Question answered | Example | |---|---|---| | Authentication | Who are you? | Login with email + password | | Authorization | What can you do? | Admin can delete; viewer can only read | | Identity | What do we know about you? | Email, roles, tenant ID in the token |

Token Types

| Type | Storage | Lifespan | Use for | |---|---|---|---| | Access token (JWT) | Memory / header | 5–60 min | API calls | | Refresh token (opaque) | HttpOnly cookie | Days–weeks | Obtain new access tokens | | Session cookie | HttpOnly cookie | Session or sliding | Traditional web apps | | API key | Server-side only | Long-lived | M2M, developer integrations |


JWT Patterns

Structure and Signing

# Python — PyJWT
import jwt
from datetime import datetime, timedelta, UTC

SECRET = "..."  # use RS256 with a key pair in production

def issue_token(user_id: str, roles: list[str]) -> str:
    return jwt.encode(
        {
            "sub": user_id,
            "roles": roles,
            "iat": datetime.now(UTC),
            "exp": datetime.now(UTC) + timedelta(minutes=15),
        },
        SECRET,
        algorithm="HS256",
    )

def verify_token(token: str) -> dict:
    return jwt.decode(token, SECRET, algorithms=["HS256"])
// TypeScript — jose
import { SignJWT, jwtVerify } from "jose";

const secret = new TextEncoder().encode(process.env.JWT_SECRET);

async function issueToken(userId: string, roles: string[]): Promise {
  return new SignJWT({ sub: userId, roles })
    .setProtectedHeader({ alg: "HS256" })
    .setIssuedAt()
    .setExpirationTime("15m")
    .sign(secret);
}

async function verifyToken(token: string) {
  const { payload } = await jwtVerify(token, secret);
  return payload;
}
// Go — golang-jwt/jwt
import (
    "github.com/golang-jwt/jwt/v5"
    "time"
)

type Claims struct {
    Roles []string `json:"roles"`
    jwt.RegisteredClaims
}

func IssueToken(userID string, roles []string, secret []byte) (string, error) {
    claims := Claims{
        Roles: roles,
        RegisteredClaims: jwt.RegisteredClaims{
            Subject:   userID,
            IssuedAt:  jwt.NewNumericDate(time.Now()),
            ExpiresAt: jwt.NewNumericDate(time.Now().Add(15 * time.Minute)),
        },
    }
    return jwt.NewWithClaims(jwt.SigningMethodHS256, claims).SignedString(secret)
}

RS256 vs HS256

| | HS256 | RS256 | |---|---|---| | Key type | Shared secret | Private/public key pair | | Who can verify | Anyone with the secret | Anyone with the public key | | Best for | Single service | Microservices, public JWKS endpoint | | Rotation | Requires coordinated redeploy | Rotate private key; publish new JWKS |

Use RS256 in production when multiple services verify tokens, or when tokens are issued by an identity provider.


OAuth2 / OIDC

Flow Selection

| Flow | Use when | Notes | |---|---|---| | Authorization Code + PKCE | Browser SPA, mobile app | No client secret on device | | Authorization Code | Server-side web app | Store client secret server-side | | Client Credentials | M2M / service accounts | No user involved | | Device Code | CLI tools, smart TVs | User authenticates on a second device |

Never use Implicit flow — it is deprecated (RFC 9700).

Authorization Code + PKCE (SPA)

// 1. Generate PKCE values
function generateCodeVerifier(): string {
  const array = new Uint8Array(32);
  crypto.getRandomValues(array);
  return btoa(String.fromCharCode(...array)).replace(/[+/=]/g, (c) =>
    ({ "+": "-", "/": "_", "=": "" })[c]!
  );
}

async function generateCodeChallenge(verifier: string): Promise {
  const data = new TextEncoder().encode(verifier);
  const digest = await crypto.subtle.digest("SHA-256", data);
  return btoa(String.fromCharCode(...new Uint8Array(digest)))
    .replace(/[+/=]/g, (c) => ({ "+": "-", "/": "_", "=": "" })[c]!);
}

// 2. Redirect to provider
const verifier = generateCodeVerifier();
sessionStorage.setItem("pkce_verifier", verifier);
const challenge = await generateCodeChallenge(verifier);

const params = new URLSearchParams({
  response_type: "code",
  client_id: CLIENT_ID,
  redirect_uri: REDIRECT_URI,
  scope: "openid profile email",
  code_challenge: challenge,
  code_challenge_method: "S256",
  state: crypto.randomUUID(), // store and verify on return
});
window.location.href = `${PROVIDER_URL}/authorize?${params}`;

// 3. Exchange code for tokens (on redirect back)
async function handleCallback(code: string): Promise {
  const verifier = sessionStorage.getItem("pkce_verifier")!;
  const res = await fetch(`${PROVIDER_URL}/token`, {
    method: "POST",
    headers: { "Content-Type": "application/x-www-form-urlencoded" },
    body: new URLSearchParams({
      grant_type: "authorization_code",
      code,
      redirect_uri: REDIRECT_URI,
      client_id: CLIENT_ID,
      code_verifier: verifier,
    }),
  });
  const { access_token, refresh_token, id_token } = await res.json();
  // store access_token in memory, refresh_token in HttpOnly cookie via backend
}

Client Credentials (M2M)

# Python — httpx
import httpx

def get_m2m_token(client_id: str, client_secret: str, token_url: str) -> str:
    r = httpx.post(
        token_url,
        data={
            "grant_type": "client_credentials",
            "client_id": client_id,
            "client_secret": client_secret,
            "scope": "api:read api:write",
        },
    )
    r.raise_for_status()
    return r.json()["access_token"]

Token Storage

| Location | XSS safe | CSRF safe | Notes | |---|---|---|---| | Memory (JS variable) | Yes | Yes | Lost on page refresh; best for SPAs | | HttpOnly cookie | Yes | No — add CSRF token | Best for refresh tokens | | localStorage | No | Yes | Never store tokens here | | sessionStorage | No | Yes | Cleared on tab close; still XSS-vulnerable |

Rule: Store access tokens in memory. Store refresh tokens in HttpOnly, SameSite=Strict cookies. Never put tokens in localStorage.


Token Refresh and Revocation

Silent Refresh Pattern

let accessToken: string | null = null;

async function getValidToken(): Promise {
  if (accessToken && !isExpiringSoon(accessToken)) return accessToken;

  const res = await fetch("/auth/refresh", {
    method: "POST",
    credentials: "include", // sends HttpOnly refresh token cookie
  });
  if (!res.ok) {
    // refresh token expired — redirect to login
    window.location.href = "/login";
    throw new Error("Session expired");
  }
  accessToken = (await res.json()).access_token;
  return accessToken;
}

function isExpiringSoon(token: string): boolean {
  const { exp } = JSON.parse(atob(token.split(".")[1]));
  return exp * 1000 - Date.now()  Callable:
    def dependency(token_data: dict = Depends(get_current_user)):
        user_roles = set(token_data.get("roles", []))
        if not user_roles.intersection(roles):
            raise HTTPException(status_code=status.HTTP_403_FORBIDDEN)
        return token_data
    return dependency

@router.delete("/users/{user_id}")
async def delete_user(
    user_id: str,
    _=Depends(require_roles("admin")),
):
    ...
// Express middleware
function requireRoles(...roles: string[]) {
  return (req: Request, res: Response, next: NextFunction) => {
    const userRoles: string[] = res.locals.user?.roles ?? [];
    if (!roles.some((r) => userRoles.includes(r))) {
      return res.status(403).json({ error: "Forbidden" });
    }
    next();
  };
}

router.delete("/users/:id", requireRoles("admin"), deleteUserHandler);

ABAC Policy Check

# Simple policy engine
def can(user: dict, action: str, resource: dict) -> bool:
    if "admin" in user["roles"]:
        return True
    if action == "read" and resource["public"]:
        return True
    if action in ("update", "delete") and resource["owner_id"] == user["id"]:
        return True
    return False

# Usage
if not can(current_user, "delete", post):
    raise HTTPException(status_code=403)

Password Hashing

Always use a slow, salted hashing algorithm. Never use MD5, SHA-1, or SHA-256 for passwords.

| Algorithm | Library (Python) | Library (Node) | Recommended? | |---|---|---|---| | bcrypt | bcrypt | bcrypt / argon2-browser | Yes | | Argon2id | argon2-cffi | argon2-browser | Yes — preferred | | scrypt | stdlib hashlib | stdlib crypto | Yes | | PBKDF2 | stdlib hashlib | stdlib crypto | Acceptable | | MD5 / SHA-* | — | — | Never |

# Python — argon2-cffi
from argon2 import PasswordHasher

ph = PasswordHasher()

def hash_password(password: str) -> str:
    return ph.hash(password)

def verify_password(hashed: str, password: str) -> bool:
    try:
        return ph.verify(hashed, password)
    except Exception:
        return False

Red Flags

  • Storing tokens in localStorage — XSS can steal every token; use HttpOnly cookies for refresh tokens and in-memory for access tokens.
  • Using HS256 across multiple services — all services share the secret; a compromise of one exposes all; use RS256 with a JWKS endpoint.
  • Long-lived access tokens — a 24-hour JWT cannot be revoked without a blacklist; keep them under 15 minutes.
  • Skipping state parameter in OAuth2 — omitting state enables CSRF attacks on the callback endpoint.
  • Rolling your own crypto — never implement JWT signing, hashing, or encryption from scratch; use audited libraries.
  • Trusting the alg header from the token — an attacker can set alg: none; always pin the algorithm server-side.
  • Returning 404 instead of 403 — security-by-obscurity doesn't prevent enumeration; return 403 Forbidden for authorization failures.
  • No token rotation on refresh — refresh tokens that never rotate are permanent credentials; rotate on every use and invalidate the old one.
  • Putting secrets in JWTs — JWTs are base64-encoded, not encrypted; any party with the token can read the payload.
  • Broad OAuth scopes — request only the minimum scopes needed; * or admin scopes violate least privilege.

Checklist

  • [ ] Access tokens expire in 15 minutes or less
  • [ ] Refresh tokens stored in HttpOnly, SameSite=Strict cookies — not localStorage
  • [ ] Access tokens stored in memory only — never persisted to storage
  • [ ] JWT algorithm pinned server-side — alg: none and algorithm-confusion attacks blocked
  • [ ] OAuth2 state parameter generated, stored, and verified on callback
  • [ ] PKCE used for all browser and mobile OAuth2 flows
  • [ ] Passwords hashed with Argon2id or bcrypt — never SHA-* or MD5
  • [ ] Refresh token rotation enabled — old token invalidated on each use
  • [ ] Role/permission check applied at the handler level, not just the route group
  • [ ] 401 returned for unauthenticated requests, 403 for unauthorized — never 404
  • [ ] Sensitive claims (PII, internal IDs) not included in JWT payload
  • [ ] HTTPS enforced on all auth endpoints — no token transmission over HTTP
  • [ ] Rate limiting applied to login, register, and token endpoints
  • [ ] Token revocation strategy documented and implemented (blacklist or short expiry)

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet — be the first.

Versions

  • v0.1.0 Imported from the upstream source.