authorization
Skill26 starsChanged 6 months agoMITView on GitHub
Use when the user writes any endpoint, server action, route handler, or query that reads or mutates user-owned data — anything taking a resource ID from the request. Catches IDOR, missing row-level checks, "logged in" mistaken for "allowed to," and role/permission gaps.
6 other projects use a version of this.
Adopted by
Most recently taken up by
- langwatch/langwatchtheir own version, first seen Oct 17, 2026
- orangehrm/orangehrmtheir own version, first seen Oct 8, 2026
- nahid-sparktales/agent-dispatchertheir own version, first seen Sep 21, 2026
- kimtth/agent-skill-100-lines-or-lesstheir own version, first seen Jun 26, 2026
- hkarpinen/agent-skillstheir own version, first seen Apr 15, 2026
- ffsshhttiikk/opencode-agents-skillstheir own version, first seen Feb 28, 2026
Other versions6 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 6
- langwatch/langwatch★ 5ksame name
- orangehrm/orangehrm★ 1.1ksame name
- nahid-sparktales/agent-dispatcher★ 59same name
- ffsshhttiikk/opencode-agents-skills★ 2same name
- kimtth/agent-skill-100-lines-or-less★ 2same name
- hkarpinen/agent-skills★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository19 other files it gives its agents.
- api-route-hardeningSkill
- auth-and-sessionsSkill
- caching-and-invalidationSkill
- database-query-safetySkill
- data-privacySkill
- error-handlingSkill
- file-uploadsSkill
- input-validationSkill
- jobs-and-queuesSkill
- llm-endpoint-safetySkill
- migrations-and-schemaSkill
- observabilitySkill
- production-readinessSkill
- reliability-patternsSkill
- rsc-boundariesSkill
- secrets-and-envSkill
- testing-strategySkill
- transactions-and-consistencySkill
- webhooksSkill
The file
We split this skill into the rules it gives an agent
1 rule. 0 are common, 1 is distinctive, and 0 are very specific to this project.
0
Common
Most of the 5 edited versions we read keep these.
1
Distinctive
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 69
If a user belongs to one or more orgs:
“Common” is measured: kept by at least 6 in 10 of the 5 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Authorization
Authentication answers "who are you." Authorization answers "are you allowed to touch this." They are not the same. Most "data leak" incidents are missing authorization, not missing authentication.
When to use
- Writing a route handler, Server Action, RPC, or RSC that takes an ID from the URL/body/params
- Building admin / team / multi-tenant features
- Adding a "share" or "transfer" feature
- Returning a list that should be scoped to the current user/tenant
- Touching
WHERE id = ?orfindByIdwith input-derived IDs
When to skip
- Public, unauthenticated, read-only endpoints with no user-scoped data
- Health checks, public marketing content
The rule, stated plainly
Every query that touches user-owned data must scope by the current user/tenant. Every. Single. One.
Logged-in does not mean allowed. Knowing the URL does not mean allowed. Being on the right page does not mean allowed.
The IDOR pattern (Insecure Direct Object Reference)
// BAD — any logged-in user can read any invoice by guessing IDs
export async function GET(req: Request, { params }: { params: { id: string } }) {
const session = await getSession()
if (!session) return new Response('Unauthorized', { status: 401 })
const invoice = await db.invoice.findUnique({ where: { id: params.id } })
return Response.json(invoice)
}
// GOOD — scoped by ownership
export async function GET(req: Request, { params }: { params: { id: string } }) {
const session = await getSession()
if (!session) return new Response('Unauthorized', { status: 401 })
const invoice = await db.invoice.findFirst({
where: { id: params.id, userId: session.userId }, // ← the load-bearing line
})
if (!invoice) return new Response('Not found', { status: 404 })
return Response.json(invoice)
}The vulnerable version "feels fine" because it has getSession(). That's authentication. It does not check whether this user owns that invoice. That's authorization.
The rules
1. Scope at the query, not after.
// BAD — fetch then check (race-prone, leaks existence, easy to forget)
const order = await db.order.findUnique({ where: { id } })
if (order.userId !== session.userId) throw new Error('forbidden')
// GOOD — scope in the WHERE clause
const order = await db.order.findFirst({
where: { id, userId: session.userId },
})2. Multi-tenant: scope by tenant, not just user.
If a user belongs to one or more orgs:
// GOOD
const project = await db.project.findFirst({
where: { id: projectId, orgId: { in: session.orgIds } },
})Better still: enforce at the database with Row-Level Security (Postgres RLS, Supabase) or a query interceptor in your ORM (Prisma extensions, Drizzle middleware). Defense in depth — a forgotten scope at the call site is still blocked at the DB.
3. 404, not 403, for "not yours."
Returning 403 Forbidden confirms the resource exists. That's an information leak. Return 404 Not Found for both "doesn't exist" and "exists but not yours" — the user should not be able to enumerate IDs.
4. Authorization checks happen on the server. Always.
Hiding a button on the client is UX, not security. Anyone can curl the endpoint. Every check happens server-side, in the route handler / Server Action / RPC.
// Client-side check is fine for UX
{ session.role === 'admin' && <DeleteButton /> }
// But the endpoint must check too
export async function DELETE(req, { params }) {
const session = await getSession()
if (session?.role !== 'admin') return new Response('Not found', { status: 404 })
// ...
}5. Permissions, not just roles, once you have more than two of them.
Roles (admin, member, viewer) work for simple apps. Once you have edge cases (a member who can manage billing but not invite users), switch to permissions:
// permission strings: "billing.manage", "members.invite", "projects.delete"
if (!session.permissions.includes('billing.manage')) {
return new Response('Not found', { status: 404 })
}A role becomes a named bundle of permissions. The check stays the same regardless of how the bundle is composed.
6. Server Actions and RSCs need authz too.
Server Actions are HTTP endpoints. RSCs run on the server but render with whatever data you fetch — if you fetch unscoped, you render someone else's data. Same rule: scope at the query.
7. Bulk operations: scope every item, not just the IDs you're given.
// BAD — trusts the IDs in the request
await db.task.deleteMany({ where: { id: { in: req.body.ids } } })
// GOOD
await db.task.deleteMany({
where: { id: { in: req.body.ids }, userId: session.userId },
})8. "Public share" links are not exemptions.
They're a separate authorization rule: "this resource has an active share token and the request presents it." Still scoped, still checked, still expirable, still revocable.
War story
A YC-batch SaaS launched a "team workspaces" feature. Every endpoint had getSession() at the top. Six months later a customer noticed they could change a number in the URL of /api/projects/123/export and download any other tenant's project. Every authenticated user on the platform could read every other tenant's data. The fix was adding orgId: session.orgId to nine findUnique calls. The cost was a public disclosure post and three churned enterprise contracts.
Quick checklist
- Every DB query with a user-supplied ID has a
userId/orgId/tenantIdclause - Authentication and authorization are both checked — never assume one implies the other
404(not403) for "exists but not yours"- Bulk endpoints scope every target row, not just the requested IDs
- Authorization runs server-side; client checks are UX only
- Postgres RLS (or ORM middleware) layered as defense in depth
- Share/invite tokens are scoped, expirable, revocable
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

