> ## Documentation Index
> Fetch the complete documentation index at: https://community.rifteo.com/llms.txt
> Use this file to discover all available pages before exploring further.
# GraphQL Raider
> Complete GraphQL security testing methodology for introspection abuse, schema enumeration, query depth and batching attacks, alias bypasses, injection through arguments, broken authorization, CSRF, WebSocket hijacking, and mass-assignment exploitation. Trigger when the user finds a /graphql endpoint, sees GraphQL queries in traffic, wants to test a GraphQL API for vulnerabilities, needs to enumerate a hidden schema without introspection, wants to bypass rate limits or brute-force credentials through GraphQL, or needs to escalate privileges through mutations.
**Status:** Stable
**Version:** 1.0.0
**Author:** Rifteo
**Tags:** graphql, pentest, api, injection, authorization, dos, idor, bug-bounty, web, websocket
**Installation**
```bash theme={"system"}
rifteo-skills add graphql-raider
```
***
## Summary
Run a full GraphQL assessment from reconnaissance to exploit validation: identify the endpoint, enumerate the schema, trigger introspection abuse or schema oracles, test batching and depth limiting, verify authorization flaws, and confirm high-impact mutation abuse with evidence.
* Fingerprints GraphQL endpoints and distinguishes server behavior from the error response alone
* Enumerates query, mutation, and subscription surfaces to reveal hidden operations and leaked object fields
* Tests GraphQL-specific DoS vectors such as depth, alias abuse, and batching that generic API testing misses
* Probes argument injection, mass assignment, IDOR, and field-level authorization issues through real mutation flows
* Produces a confirmed exploit chain only when the evidence supports it
***
## SKILL.md file
### GraphQL Raider
Test GraphQL endpoints as a high-risk API surface by moving from reconnaissance to exploit-chain validation without guessing.
#### What Does It Check?
The skill covers the GraphQL-specific attack surface: introspection abuse, field suggestion oracles, depth and batching abuse, alias-based rate-limit bypass, arguments that can carry injection payloads, broken authorization on mutations, CSRF via GET-style mutations, and subscription or mass-assignment abuse.
**In scope:**
* `/graphql` and common GraphQL endpoint variants
* Introspection and schema enumeration
* Query depth, batching, alias, and rate-limit bypass checks
* Injection via arguments (SQLi, NoSQLi, SSTI)
* Broken object-level and field-level authorization
* CSRF via GET mutations and mutation abuse through GraphQL
* Subscription hijacking and mass assignment via mutation parameters
* WAF and protection bypass patterns
**Out of scope:**
* Generic web app testing outside the GraphQL endpoint
* Full pentest reporting across a whole application; this skill targets the GraphQL surface specifically
#### How It Works
**Phase 1 — Reconnaissance: Find & Fingerprint GraphQL**
Check common GraphQL paths and send a minimal `__typename` request to confirm the endpoint and server type. Response signatures reveal whether introspection is enabled or if the server is exposing a particular framework behavior.
**Phase 2 — Schema Enumeration**
Enumerate the schema, list types, and prioritize mutation surfaces. The skill extracts the hidden objects, arguments, and sensitive fields most likely to produce privilege escalation or data leakage.
**Phase 3 — Attack Selection**
Move through the highest-value GraphQL abuse classes in order: introspection abuse, field suggestion oracle, query depth attacks, batching, alias-based rate-limit bypass, injection through arguments, broken object-level auth, broken field-level auth, CSRF via GET mutations, circular fragment DoS, subscription hijacking, and mass assignment.
**Phase 4 — Confirm the Exploit**
Only treat a finding as real if it is supported by a reproducible response or a working mutation. The skill validates privilege escalation or data access before writing up the result.
**Phase 5 — Report the Finding**
Write a structured result with the endpoint affected, the root cause, the exploit path, relevant evidence, and a concrete remediation recommendation.
#### Output
| Status | Condition |
| --------------------- | -------------------------------------------------------------------- |
| Confirmed finding | A GraphQL bug was reproduced and the impact is validated |
| Blocked but promising | A vulnerability is likely but needs one missing condition to confirm |
| No issue found | The endpoint rejects the payload or the behavior is not exploitable |
Example output structure:
```text theme={"system"}
Title: GraphQL mutation allows privilege escalation via updateUser
Severity: Critical
Affected endpoint: POST /graphql
Steps to reproduce:
1. Authenticate as a low-privilege user.
2. Submit a mutation that changes another user's role or admin flag.
3. Observe the response returns the modified user record.
Impact: Any authenticated user can promote another account to admin.
Evidence: [request payload] / [response snippet]
Recommendation: Enforce object-level authorization on every mutation and reject any update to privileged fields unless the caller has explicit admin rights.
```
#### Known Limitations
* The skill requires a real target or a reproducible GraphQL payload to validate behavior
* Some protections depend on server-side API logic and may require account setup, multi-user testing, or WebSocket access
* It does not replace a full application pentest; it focuses on the GraphQL attack surface
***
## Benchmark Results
Tested on claude-sonnet-4-6 via Claude Code CLI. Same prompt, same model, same target. The only variable is whether the skill is loaded.
| Metric | Without Skill | With Skill |
| ----------------- | --------------- | ---------------- |
| Turns to complete | 3 | 1 |
| Total tokens | \~8,200 | \~2,400 |
| Time | 185s | 52s |
| Attack coverage | 2/14 techniques | 14/14 techniques |
| False positives | 3 unverified | 0 |
This benchmark reflects a real GraphQL target where the agent had to discover and exploit a schema-backed attack chain. Without the skill, the agent guessed at a mutation, stopped at partial enumeration, and left the attack path incomplete. With `graphql-raider`, the workflow covered the full GraphQL methodology in a single turn and produced a validated exploit chain with zero false positives.
***
## Related skills
Systematic IDOR/BOLA detection methodology with recon, testing, and bypass techniques
Full JWT attack methodology including alg confusion, weak secret brute-force, and header exploitation
Convert raw pentest notes and exploit output into structured, report-ready findings