Works with the AI Agents you already use
OWASP Sentinel
Turn your AI agent into a disciplined web & API security tester — OWASP Top 10:2025 and API Top 10:2023, with a client-ready CVSS report.
$9
OWASP Sentinel
Example session with this skill installed
Run a security assessment of our ShopNova staging API at https://staging-api.shopnova.example.
- Authorization: I'm the CTO and approve this test (ref. SEC-2026-031)
- Scope: staging API only — payment sandbox, email delivery and DoS testing are out of scope
- Test accounts:
customerAandcustomerB(role: customer),adminQA(role: admin) - Rate limit: max 5 requests/second
- Spec: OpenAPI 3.0 at
/openapi.json
Give me a full report with CVSS scores and how to fix each issue.
- Read your context and instructions
- Compiled the owasp sentinel
Authorization confirmed (ref. SEC-2026-031), scope locked to staging-api.shopnova.example, rate limit set to ≤ 5 req/s. I mapped
23 endpoints from the OpenAPI spec plus 2 undocumented ones, then ran the full OWASP test catalogue with the three test accounts.
Result: overall risk High — 6 findings (2 High, 2 Medium, 1 Low, 1 Informational). The critical problem is access control: any customer can read other customers' orders and promote themselves to admin. Full report below.
OWASP Sentinel — Security report: ShopNova API (2026-10-15)
Executive summary
Overall risk: High. The assessment of the ShopNova staging API identified
6 findings: 0 Critical, 2 High, 2 Medium, 1 Low, 1 Informational.
The most serious issues are in
access control: any logged-in customer can read and modify other customers' orders, and can grant themselves administrator rights by editing their own profile.
Top 3 priority actions
- Enforce object-level ownership checks on every
/ordersendpoint (ID-01). - Use an allow-list of writable fields on
PATCH /users/meand never acceptrolefrom clients (ID-02). - Add rate limiting and progressive lockout on
/auth/login(ID-04).
Scope and methodology
| Item | Detail |
|---|---|
| Target | https://staging-api.shopnova.example (REST, JSON) |
| Environment | Staging (production out of scope) |
| Window | 2026-10-14 09:00 – 2026-10-15 18:00 CEST |
| Authorization | Written approval from ShopNova CTO, ref. SEC-2026-031 |
| Accounts | customerA, customerB (role: customer), adminQA (role: admin) |
| Rate limit agreed | ≤ 5 req/s |
| Inputs | OpenAPI 3.0 spec (v1), frontend at https://staging.shopnova.example |
| Exclusions | Payment provider sandbox, email delivery, DoS testing |
Frameworks: OWASP Top 10:2025, OWASP API Security Top 10:2023, OWASP WSTG.
Limitations: black/grey-box, no source code access, non-destructive testing only.
Coverage matrix
| OWASP category | Tested | Result |
|---|---|---|
| A01 Broken Access Control / API1, API3, API5 | Yes | ❌ ID-01, ID-02 |
| A02 Security Misconfiguration / API8 | Yes | ⚠️ ID-05 |
| A03 Software Supply Chain Failures / API10 | Yes | ✅ No issues |
| A04 Cryptographic Failures | Yes | ⚠️ ID-03 |
| A05 Injection | Yes | ✅ No issues |
| A06 Insecure Design / API6 | Yes | ✅ No issues |
| A07 Authentication Failures / API2 | Yes | ⚠️ ID-03, ID-04 |
| A08 Software & Data Integrity Failures | Yes | ✅ No issues |
| A09 Logging & Alerting Failures | Partial | ℹ️ Client confirmed alerts fired for login attempts |
| A10 Mishandling of Exceptional Conditions | Yes | ✅ No issues |
| API4 Unrestricted Resource Consumption | Yes | ⚠️ ID-04 |
| API7 SSRF | Yes | ✅ Avatar URL import uses an allow-list |
| API9 Improper Inventory Management | Yes | ℹ️ ID-06 |
Findings
[ID-01] Customers can read and modify other customers' orders (BOLA)
- Severity: High — CVSS 7.1 (
CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:L/A:N) - Status: Confirmed
- Category: A01:2025 / API1:2023 — CWE-639
- Endpoint(s):
GET /api/v1/orders/{id},PATCH /api/v1/orders/{id}
Description: The API checks that the request carries a valid token but not that the order belongs to the token's owner. Order IDs are sequential integers, so they are trivial to guess.
- Steps to reproduce:
- Log in as
customerBand create order10482. - Log in as
customerA. - Send
GET /api/v1/orders/10482withcustomerA's token →200 OKwithcustomerB's order. - Send
PATCH /api/v1/orders/10482with{"shippingAddress": {...}}→200 OK, address changed.
- Log in as
- Evidence:
GET /api/v1/orders/10482 HTTP/1.1 Host: staging-api.shopnova.example Authorization: Bearer eyJhbGciOi...[REDACTED] # customerA HTTP/1.1 200 OK {"id":10482,"customerId":"cus_B…[REDACTED]","items":[…],"shippingAddress":{"name":"[REDACTED]", …}} - Impact: Exposure of every customer's name, address and purchase history; ability to redirect deliveries.
Remediation: Resolve orders through the authenticated user (WHERE id = ? AND customer_id = :currentUser) or a central authorization policy applied to every object-level route. Consider non-sequential identifiers as defence in depth. ASVS V4.2.1.
- References: OWASP API1:2023, CWE-639.
[ID-02] Privilege escalation via mass assignment on profile update
- Severity: High — CVSS 8.8 (
CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H) - Status: Confirmed
- Category: A01:2025 / API3:2023 — CWE-915
- Endpoint(s):
PATCH /api/v1/users/me
Description: The endpoint binds the whole request body to the user model. The role property, not documented in the spec, is accepted and persisted.
- Steps to reproduce:
- Log in as
customerA. - Send
PATCH /api/v1/users/mewith{"displayName":"A","role":"admin"}→200 OK. - Send
GET /api/v1/admin/users→200 OK(previously403). - Role reverted to
customerwithadminQAafter confirmation.
- Log in as
- Evidence:
PATCH /api/v1/users/me HTTP/1.1 Authorization: Bearer eyJhbGciOi...[REDACTED] Content-Type: application/json {"displayName":"A","role":"admin"} HTTP/1.1 200 OK {"id":"usr_A…","displayName":"A","role":"admin"} - Impact: Any customer can take over the platform's administrative functions.
Remediation: Use an explicit DTO / allow-list of client-writable fields (displayName, phone, preferences). Manage roles only through a dedicated admin endpoint protected by function-level authorization. ASVS V5.1.2, V4.1.3.
- References: OWASP API3:2023, CWE-915.
[ID-03] Long-lived JWTs are not revoked on logout
- Severity: Medium — CVSS 4.8 (
CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:L/I:L/A:N) - Status: Confirmed
- Category: A07:2025 / API2:2023 — CWE-613
- Endpoint(s):
POST /api/v1/auth/logout, all authenticated endpoints
Description: Access tokens are valid for 30 days (exp claim) and remain accepted after logout and after a password change. Signature verification and algorithm handling were correct (alg: none and modified claims were rejected).
- Steps to reproduce:
- Log in as
customerA, store the token. - Call
POST /auth/logout→204. - Reuse the stored token on
GET /users/me→200 OK.
- Log in as
Impact: A stolen token (e.g. from a shared device or log) gives access for up to 30 days, and users cannot cut it off.
Remediation: Short-lived access tokens (≤ 15 min) with rotating refresh tokens; server-side revocation (deny-list or token version) on logout and password change. ASVS V3.3.1, V3.5.2.
- References: OWASP API2:2023, CWE-613.
[ID-04] No rate limiting on login
- Severity: Medium — CVSS 5.3 (
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:N/A:N) - Status: Confirmed
- Category: A07:2025 / API4:2023 — CWE-307
- Endpoint(s):
POST /api/v1/auth/login
Description: 10 consecutive failed logins against customerA within 5 seconds all returned 401 with no throttling, lockout, CAPTCHA or RateLimit-* headers. The 11th attempt with the correct password succeeded.
- Impact: Enables credential stuffing and password guessing at scale.
Remediation: Per-account and per-IP rate limiting with exponential back-off, alerting on spikes, and optional CAPTCHA after N failures. ASVS V2.2.1.
- References: OWASP API4:2023, CWE-307.
[ID-05] Missing security headers on the web frontend
- Severity: Low — CVSS 3.1 (
CVSS:3.1/AV:N/AC:H/PR:N/UI:R/S:U/C:L/I:N/A:N) - Status: Confirmed
- Category: A02:2025 — CWE-693
- Endpoint(s):
https://staging.shopnova.example/*
Description: Responses lack Content-Security-Policy and Strict-Transport-Security. No XSS was found, but there is no second line of defence if one is introduced.
Remediation: Add a restrictive CSP (start in Report-Only), Strict-Transport-Security: max-age=31536000; includeSubDomains, and frame-ancestors 'none'. ASVS V14.4.
- References: OWASP Secure Headers Project, CWE-693.
[ID-06] Legacy API version and Swagger UI exposed
- Severity: Informational — CVSS 0.0
- Status: Informational
- Category: API9:2023 / A02:2025 — CWE-1059
- Endpoint(s):
/api/v0/*,/docs
Description: /api/v0 responds but is not in the current spec; /docs serves Swagger UI. Spot checks on /api/v0/orders/{id} showed the same ownership checks as v1 are
missing there too (covered by ID-01). Acceptable on staging, but confirm neither is reachable in production.
Remediation: Decommission v0 or apply the same controls and document it; restrict /docs to internal networks in production.
- References: OWASP API9:2023.
General recommendations
- Centralize authorization in one middleware/policy layer and add automated tests that run every endpoint with a second user's token.
- Add an API contract test that fails when a response contains fields not in the spec.
- Re-test ID-01 and ID-02 after fixes before the next release.
Activity log (appendix)
| Time (CEST) | Method | Endpoint | Purpose |
|---|---|---|---|
| 10-14 09:12 | GET | /openapi.json | Surface mapping |
| 10-14 09:40 | GET | /api/v1/orders/10482 | BOLA test (customerA → customerB) |
| 10-14 09:42 | PATCH | /api/v1/orders/10482 | BOLA write test |
| 10-14 10:05 | PATCH | /api/v1/users/me | Mass assignment test |
| 10-14 10:07 | GET | /api/v1/admin/users | Confirm escalation |
| 10-14 10:09 | PATCH | /api/v1/admin/users/usr_A… | Revert role (adminQA) |
| 10-14 11:30 | POST | /api/v1/auth/login ×10 | Lockout check |
| 10-14 14:20 | POST | /api/v1/auth/logout | Token revocation check |
| 10-15 09:15 | GET | /api/v0/orders/10482 | Inventory check |
| … | … | … | … |
Next steps
Fix ID-01 and ID-02 first — both are exploitable by any registered customer. I can draft the ownership-check middleware and the profile-update allow-list for your stack.
2. Plan ID-03 and ID-04 for the next sprint (short-lived tokens + revocation, login rate limiting).
3.
Re-test: once fixes are deployed to staging, ask me to re-run the assessment on the affected endpoints and I'll append a retest section to this report.
Connects securely to your tools. The creator never sees your data.
About this skill
OWASP Sentinel guides your AI agent through an authorized, non-destructive security assessment of a deployed web application or REST API, and delivers a professional Markdown report with CVSS scores, CWE mapping, redacted evidence and actionable fixes.
How it works:
- Authorization gate: the agent won't send a single request until permission, scope, test accounts and rate limits are confirmed.
- Surface mapping: endpoints, auth mechanism, API versions and sensitive business flows.
- OWASP test catalogue: 13 test families covering every OWASP Top 10:2025 and API Security Top 10:2023 category, including BOLA/IDOR, BFLA, mass assignment, JWT weaknesses, injection, SSRF, rate limiting and business-logic flaws.
- Validation: each finding is reproduced, false positives are ruled out, and it is classified as Confirmed, Likely or Informational.
- Report: executive summary, coverage matrix, CVSS v3.1 vectors, remediation with ASVS references and an activity log.
Safe by design: benign probe markers, request-rate limits, minimal proof of concept, and automatic stop on service degradation or real personal data. No scripts, no binaries, no exploits: just a structured methodology your agent follows step by step.
Stack-agnostic: works on any backend (Node, Java, Python, .NET, PHP, Go…) because it tests HTTP behaviour, not code.
Includes a full sample report on a fictional e-commerce API.
Requires explicit written authorization and scope confirmation before testing. Does not support SAST, DoS, or physical security testing. Restricted to non-destructive confirmation techniques.
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 26 days ago