More screenshots
Works with the AI Agents you already use
Identity Federation: SAML, OIDC, OAuth2 Review
Audit SSO the way attackers map it: draw the full SP–IdP trust chain, test SAML assertion integrity, OIDC flow checks, and the session layer, and verify every finding by reproduction.
$5
Identity Federation: SAML, OIDC, OAuth2 Review
Example session with this skill installed
We sell to enterprises and offer SAML SSO plus OIDC for smaller customers. A customer's security team is about to pen-test our login. Can you audit our federation setup first, the way an attacker would map it?
- Read your context and instructions
- Compiled the identity federation
Step 1, the chain: user → your SP (login initiated) → IdP (authenticates) → SAML assertion → your SP (validates, creates session). Trust assumptions written on each arrow; the unstated one — "the SP probably verifies the assertion itself" — became the first test target.
Step 2, captures: full proxy-captured logins for a standard user and a tenant admin in your test tenant, saved as baselines. Every mutation below is a diff against these.
Step 3, link tests. SAML integrity: the response wrapper is signed, but replaying the flow with the assertion's signature element removed still creates a session. Escalating only inside the test tenant: the assertion's NameID is changed to the tenant admin test account — unsigned — and the SP creates an admin session. Assertion-level signature verification is not enforced. Audience and expiry checks, by contrast, hold: a cross-SP assertion and an expired one are both rejected, recorded as negative evidence.
OIDC side: redirect_uri is exact-match, state and nonce are enforced, codes are single-use with PKCE on the public client — cleared with captures. Session layer: SP logout leaves the IdP session alive, so a "logged out" user is silently re-authenticated on the next login attempt.
Step 4, verdicts: unsigned-assertion acceptance rated Critical (reachable by anyone who can reach the login endpoint; fix = reject assertions lacking a valid assertion-level signature before session creation, plus a regression test replaying this capture). Orphaned IdP session on logout rated Low with the capture attached.
Connects securely to your tools. The creator never sees your data.
About this skill
Single sign-on breaks at the links, not the endpoints — an assertion the SP never really verified, a redirect_uri matched by prefix, a logout that leaves the IdP session alive. This skill is the audit routine for the federation layer. It starts by drawing the complete trust chain (user → SP → IdP → token → SP session) with the trust assumption written on every arrow, then captures a real login per role as the lab every test replays and mutates. The checks cover the classic link failures: SAML signature coverage (assertion vs. response wrapper), algorithm downgrade, audience/recipient mismatch, and replay windows; OIDC/OAuth2 redirect_uri strictness, state and nonce enforcement, PKCE, code single-use, and multi-tenant issuer confusion; plus the session layer — logout completeness in both directions, session fixation, and claim-to-role mapping. Nothing is reported on theory: every finding carries the exact mutated request, the response proving the bypass, an impact statement, and a rating from a trigger-requirement table. Includes a finding template and the full usual-suspects checklist.
- Encrypted assertions can't be mutated without a test configuration; the skill records those checks as blocked rather than passed when the IdP admin can't provide one.
- Some IdP tenant settings (Okta, Entra ID, ADFS) need the IdP admin's cooperation to verify — the audit flags exactly which items need their eyes.
- It covers the federation layer only; JWT/API-token issues on non-federated APIs belong to API security testing.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 24 days ago
Permissions
No special permissions declared or detected