appsec.fyi

Server-Side Template Injection (SSTI) Resources

Post Share

A curated AppSec resource library covering XSS, SQLi, SSRF, IDOR, RCE, XXE, OSINT, and more.

Server-Side Template Injection (SSTI)

Server-Side Template Injection (SSTI) occurs when user input is concatenated directly into a server-side template instead of being passed in as a parameter. Because templating engines are designed to evaluate expressions and call methods, an attacker who controls part of the template can frequently escalate from a reflected math expression like {{7*7}} all the way to remote code execution in the language the engine runs on.

The vulnerability class was first formalized by James Kettle's PortSwigger research, which demonstrated generic exploits and sandbox escapes across the most popular template engines. Since then, real-world SSTI has been found in Jinja2 (Python/Flask), Twig (PHP), Freemarker and Velocity (Java), Smarty and Mako, and even nominally sandboxed environments — sandbox escapes for Jinja2 are republished almost every year, and Apache Camel, OpenMetadata, and Alfresco have all shipped critical SSTI CVEs in production frameworks.

SSTI is easy to confuse with XSS at first glance because both involve injection into rendered output, but the impact is fundamentally different. XSS executes in the victim's browser; SSTI executes on the server, often as the application user, and leads directly to file read, environment access, and full RCE. Detection starts with simple expression probes ({{7*7}}, ${7*7}, *{7*7}), then engine fingerprinting through error messages, then engine-specific gadget chains.

This page collects PortSwigger's foundational research, Web Security Academy labs, HackTricks and PayloadsAllTheThings references, the standard tooling (tplmap, SSTImap), and engine-specific exploitation guides and CVE writeups for Jinja2, Twig, Freemarker, and beyond.

From PortSwigger Research

Template injection turns a formatting feature into an interpreter

SSTI happens when user input is concatenated into a template rather than passed to one as data. The distinction is easy to miss because both look like string handling, and the consequence is disproportionate: template engines are not text substituters, they are expression evaluators with access to the language's object model. Rendering a template containing attacker text means evaluating attacker expressions.

The feature most exploited is not a flaw at all but reflection. In Python's Jinja2 the standard path runs through object attributes — reaching the base object type, enumerating its subclasses, finding one that exposes a useful import or process call — which is why the payloads look like attribute chains rather than code. Java engines such as Freemarker and Velocity expose class loading and can reach runtime execution. PHP's Twig and Smarty offer filters and registered functions. The mechanism differs, and the shape of the attack does not: get from an evaluated expression to something that touches the host.

Detection begins with arithmetic. If {{7*7}} renders as 49, the input is being evaluated rather than displayed, and that single observation separates SSTI from reflected XSS, which is what these inputs are usually mistaken for. From there the syntax that errors versus evaluates fingerprints the engine, and the engine determines the escape path.

Sandboxes exist in several of these engines and should be treated as a delay rather than a boundary; Jinja2's in particular has been escaped repeatedly, and the escapes typically use ordinary language features rather than engine bugs. Logic-less templating removes the class outright, which is the strongest argument for it.

Where this appears in practice: email and notification templating, customizable reports and invoices, white-label theming, admin-editable content, and anything with a "insert a variable here" feature. Those are all legitimate template-editing features, which makes deciding whether the capability is intended part of the finding.

Start here: the SSTI learning pathAn ordered route through this library — orientation, then methodology, then research. Read the SSTI guideA long-form, source-cited deep dive synthesized from every resource below. The comprehensive SSTI guide on chs.usA hand-written, in-depth practitioner guide — attacks, testing, and prevention.
Date Added Link Excerpt
2026-11-19 NEW 2026NocoDB server-side template injection intermediate RCENocoDB versions prior to 2026.09.0 are vulnerable to server-side template injection. This allows attackers to execute arbitrary code on the server by submitting specially crafted input. The vulnerability stems from improper handling of user-supplied data within the template rendering process. Successful exploitation could lead to unauthorized access, data breaches, or complete server compromise. Users are advised to update to a patched version to mitigate this risk.
2026-11-19 NEW 2026Nuxt UI untrusted markdown to RCE via Vue SSR template compilation intermediate RCENuxt UI, when used with Comark or MDC for server-side markdown rendering, contains a vulnerability. This flaw allows for arbitrary command execution by manipulating the markdown input. Attackers can exploit this to achieve Remote Code Execution (RCE) through Vue SSR template compilation.
2026-11-18 NEW 2026Exploiting SSTI in BugForge's 2026 Vibe Predictor intermediateThis content details the exploitation of Server-Side Template Injection (SSTI) vulnerabilities within BugForge's 2026 Vibe Predictor challenge. The author approached the task with a professional testing methodology. The summary focuses on the core vulnerability and the application being tested, without mentioning any bounty details as none were provided.
2026-11-18 NEW 2026High severity SSTI advisory newsAdvisory for CVE-2026-21596, a high-severity Server-Side Template Injection (SSTI) vulnerability in Atlassian Bamboo Data Center versions 12.0.0 and 12.1.0. An authenticated attacker can exploit this to execute arbitrary commands, impacting confidentiality, integrity, and availability. Atlassian advises upgrading to fixed versions 12.1.13 or later.
2026-11-18 NEW 2026LetMeOut: a three-letter Jinja jail escape news 1 min read PythonLibrary for exploiting Jinja2 sandbox jail vulnerabilities. This library demonstrates how to bypass the sandbox's restrictions, which limit template length to 170 characters and allow only three distinct letters, by utilizing fullwidth Unicode characters like U+FF3F to trick the filter. It reveals techniques for accessing `__globals__` and ultimately `breakpoint()` to achieve remote code execution, as seen in the hiCTF challenge.
2026-11-17 NEW 2026Twig PHP code injection via {% use %} template name intermediate RCEWriteup detailing CVE-2026-46633, a PHP code injection vulnerability in Twig. The vulnerability arises from improper escaping of single quotes in template names used within {% use %} tags, allowing arbitrary PHP code execution by terminating string literals in compiled cache files. This bypasses the Twig sandbox and is reachable even from sandboxed templates. → advisories.gitlab.com
2026-11-17 NEW 2026DataGerry SSTI via DocAPI templates (CVE-2026-57853) news 3 min readWriteup of CVE-2026-57853 details a critical server-side template injection vulnerability in DataGerry versions prior to 3.3.0. Authenticated attackers with `base.docapi.template.add` permission can inject arbitrary Jinja2 payloads via the `template_data` field. This payload is later rendered in a non-sandboxed environment, allowing exploitation of the `cycler.init.globals.os` gadget chain for arbitrary operating system command execution as the DataGerry web process user, potentially achieving root access in default Docker deployments.
2026-11-17 NEW 2026Batflat CMS SSTI in /contact endpoint intermediate RCEBatflat CMS has a Server-Side Template Injection (SSTI) vulnerability in its `/contact` endpoint. The CMS fails to properly sanitize user-supplied data before it's processed by the template engine. This weakness allows attackers to inject malicious template code, potentially leading to remote code execution on the server. This flaw impacts the security of websites using Batflat CMS.
2026-11-17 NEW 2026Secure Code Review Challenge #9: FreeMarker SSTI intermediate 16 min read RCELibrary for analyzing Spring Boot applications, this entry details a FreeMarker Server-Side Template Injection (SSTI) vulnerability in a project management tool. It walks through the review process, identifying entry points, dangerous sinks like the customizable status report template, and exploit development. The analysis highlights how FreeMarker's templating engine, combined with inadequate input sanitization, allows a low-privileged member to execute shell commands on the server, even with existing security measures like JWT authentication and CSRF tokens in place.
2026-11-17 NEW 2026CTF@CIT 2026 Temporary Destruction: Jinja SSTI blacklist bypass advanced 1 min read Python RCEWriteup of Temporary Destruction, a Jinja SSTI vulnerability. The application directly renders user input via `render_template_string`, attempting to block dunder methods with a regex. This blacklist is bypassed using escaped underscores (`\x5f`) in conjunction with Jinja's attribute access. The exploit leverages this to access `os.popen` and read the `/tmp/flag.txt` file, achieving remote code execution.
2026-11-16 NEW 2026Studying an SSTI challenge intermediate 3 min readWriteup detailing an SSTI challenge involving a WAF and a Flask application using Jinja2. The challenge bypasses a WAF by exploiting how `aiohttp` dechunks HTTP payloads before processing, allowing the injection of characters like `{{` and `}}` that the WAF inspects on raw wire bytes but are filtered later by the application. The writeup touches on static analysis, reverse engineering with IDA, blacklists, and payload splitting techniques.
2026-11-16 NEW 2026boro CTF 2026 – NERV challenge writeup intermediate 9 min readWriteup of a Server-Side Template Injection (SSTI) vulnerability in the NERV challenge from boro CTF 2026, exploiting a Jinja2 `render_template_string` flaw. The challenge involved a Flask application where user input was directly concatenated into a template without proper sanitization, leading to arbitrary command execution via Python object traversal. The writeup details reconnaissance, identifying rabbit holes, confirming the SSTI with `{{7*7}}`, and escalating to RCE using payloads like `{{ cycler.__init__.__globals__.os.popen('id').read() }}` to extract the flag.
2026-11-16 NEW 2026H7CTF 2026 Quals – Merged challenge writeup intermediate 6 min readWriteup detailing a Server-Side Template Injection (SSTI) vulnerability in Certmarq, a certificate-issuing web application, discovered during the H7CTF 2026 Quals. A superficial content filter blocking the literal string `__` can be bypassed by dynamically constructing the attribute name `__globals__` using Jinja's `~` operator and `attr` filter, ultimately leading to access to server-side Python objects and command execution.
2026-11-16 NEW 2026Formie missing authorization leads to SSTI intermediate 6 min read AuthZLibrary for Craft CMS and Formie, addressing a critical missing authorization vulnerability (GHSA-wm3q-jh2w-mm72) in the `EmailController`. This flaw allows any authenticated user to trigger server-side template injection (SSTI) by sending spoofed emails or rendering arbitrary notification content through unpermissioned actions like `send-test-email` and `preview`. The vulnerability stems from the `EmailController` not extending `SettingsAccessController`, thus bypassing authorization checks and allowing attacker-controlled input to be rendered as Twig templates.
2026-11-16 NEW 2026Vvveb incomplete fix: PHP injection into server-side templates intermediate RCEVvveb versions up to 1.0.8.6 suffer from an incomplete fix for a PHP injection vulnerability in server-side templates. This flaw allows for the execution of arbitrary PHP code on the server. Users should update to a patched version to mitigate this security risk.
2026-11-16 NEW 2026From a broken error message to full server RCE: Jinja2 SSTI writeup ($9,000) intermediate Bug Bounty Python RCEA security researcher discovered a Server-Side Template Injection (SSTI) vulnerability in Jinja2, leading to full Remote Code Execution (RCE) on the server. The exploit began with a seemingly minor error message, which upon closer inspection, revealed exploitable information. This vulnerability allowed the attacker to execute arbitrary code. The researcher was awarded a $9,000 bounty for their findings.
2026-11-15 NEW 2026Twig: PHP code injection via crafted {% use %} template names intermediate RCEAdvisory for GHSA-25fx-hhx2-xwq3 detailing a Twig Server-Side Template Injection (SSTI) vulnerability in Thelia 3.0.0. Authenticated users with "Messages" permission can exploit this flaw by crafting {% use %} template names to execute arbitrary PHP code, read sensitive files, or write to the filesystem, potentially leading to full system compromise. This issue has been resolved in thelia/twig-engine-module 1.0.9.
2026-11-15 NEW 2026CWE-1336 template injection tracker: recent SSTI CVEs (FortiManager, mailcow, OFBiz) intermediate 10 min readLibrary tracking recent server-side template injection (SSTI) CVEs, detailing vulnerabilities in FortiManager, mailcow, Apache OFBiz, Craft CMS, LMFDB, Quest Foglight, Amazon Virtual Engineering Workbench, Frappe Framework, and Bricks Theme. These instances of CWE-1336 involve improper neutralization of template engine syntax, leading to code execution with varying privilege levels and impacts, from restricted administrator accounts to SYSTEM privileges and cross-tenant data access.
2026-11-15 NEW 2026CVE-2026-107248 (CVSS 8.8) template injection news 2 min readWriteup of CVE-2026-107248, a high-severity server-side template injection (SSTI) vulnerability in Thelia 3.0.0's Email/Messages feature. This flaw, exploitable by authenticated users with "Messages" permission, allows the execution of unsandboxed Twig templates, enabling arbitrary PHP function execution, sensitive file access, and filesystem writes. Exploitation can lead to full application and host compromise. The vulnerability is patched in thelia/twig-engine-module 1.0.9 and is associated with CWE-94 and CWE-1336.
2026-11-15 NEW 2026SSTI: detection, exploitation, and payloads beginner RCELibrary for detecting and exploiting Server-Side Template Injection (SSTI) vulnerabilities in Jinja2, Twig, Freemarker, and ERB. It details detection methods using mathematical expressions, fingerprinting techniques to identify specific template engines, and provides concrete RCE payloads for Python/Flask, PHP, Java, and Ruby. The library also includes the SSTImap tool for automating SSTI discovery and offers prevention strategies like using `render_template` over `render_template_string`, employing Jinja2's `SandboxedEnvironment`, and input validation.
2026-11-14 NEW 2026SSTI: fingerprinting the template engine and getting RCE intermediate 2 min read RCELibrary for detecting and exploiting Server-Side Template Injection (SSTI) vulnerabilities. It details methods for fingerprinting template engines like Jinja2, Twig, FreeMarker, ERB, Mako, and Thymeleaf, including techniques for bypassing filters and achieving remote code execution (RCE) via commands like `id` or reverse shells. The library highlights tools such as SSTImap and references the PayloadsAllTheThings list for broader coverage, while distinguishing SSTI from client-side template injection (XSS).
2026-11-14 NEW 2026SSTI Bug Bounty 2026: server-side template injection to RCE on 5 template engines intermediate 24 min read Bug Bounty RCELibrary for detecting and exploiting Server-Side Template Injection (SSTI) vulnerabilities. It details engine-specific exploitation chains for Jinja2, Twig, Freemarker, Pebble, and Velocity, and covers techniques like polyglot probes, arithmetic confirmation, and out-of-band callbacks for discovery, enabling researchers to find critical RCE findings in bug bounty programs.
2026-11-13 2026DB-GPT Jinja2 skill template SSTI (CVE-2026-123020) intermediate AI RCEWriteup of CVE-2026-123020 impacting DB-GPT's Jinja2 Template Handler. This Server-Side Template Injection (SSTI) vulnerability in `FileBasedSkill.get_prompt` arises from improper neutralization of special template elements, allowing remote exploitation. The issue is resolved in DB-GPT version 0.8.2, with patch identifier `a9e7a3c026d114fcfb4ffe3eaf78851a636ded16`.
2026-11-13 2026DataGerry 3.3.0 SSTI via DocAPI template rendering (CVE-2026-57853) intermediate Python RCEWriteup of CVE-2026-57853 in DataGerry, a server-side template injection vulnerability exploitable via the DocAPI. Authenticated attackers with `base.docapi.template.add` permission can inject arbitrary Jinja2 payloads through POST /rest/docapi/template. These payloads are rendered un-sandboxed when a user calls GET /rest/docapi/template/<id>/render/<object_id>, allowing exploitation of the `cycler.__init__.__globals__.os` gadget chain for remote code execution as the web process user, potentially root. → vulncheck.com
2026-11-12 2026Template engine injection (SSTI) advisory newsThe advisory details a Server-Side Template Injection (SSTI) vulnerability in a template engine. This flaw arises from the improper neutralization of special elements within the engine's processing, allowing attackers to inject malicious code. Successful exploitation could lead to arbitrary code execution, potentially compromising the server. The advisory highlights the severity of this vulnerability and recommends immediate patching or mitigation strategies to prevent exploitation.
2026-11-12 2026Template engine injection (SSTI) advisory newsAdvisory for Fortinet FortiManager versions 8.0.0 through 7.2, and FortiManager Cloud versions 8.0.0 through 7.2. This vulnerability, GHSA-p5xc-69mw-v3hr, involves improper neutralization of special elements used in a template engine, potentially allowing a low-privileged Device Manager admin to achieve super admin privileges on managed FortiGates through command injection.
2026-11-12 2026CVE-2026-46634: Twig template_from_string sandbox bypass news 1 min readWriteup of CVE-2026-46634 details a Twig template sandbox bypass, scored CVSS 9.8, affecting versions of Twig from 3.9.0 up to 3.26.0. This vulnerability, a protection mechanism failure (CWE-693), occurs when a sandbox is enabled via a source policy and a sandboxed template calls `template_from_string` or includes other templates. Mitigation involves upgrading to Twig 3.26.0, removing `template_from_string` from sandbox allowed functions, or disabling the `StringLoaderExtension` entirely when using a sandbox.
2026-11-12 202630 minutes to root: how AI found an RCE (Nunjucks SSTI) in local government software intermediate AI RCEAn AI agent, guided by a researcher, discovered a critical Remote Code Execution (RCE) vulnerability in OpenStad, a local government software, within 30 minutes. The vulnerability, identified as a Nunjucks Server-Side Template Injection (SSTI), allows authenticated attackers to gain root access. This finding highlights a significant security risk for any organization utilizing OpenStad and underscores the potential of AI in accelerating security research.
2026-11-12 2026Opik SDKs render Jinja2/Nunjucks prompt templates without a sandbox intermediate 1 min read AI Python RCELibrary advisories detail GHSA-5mc6-2qfj-59cg, where Opik SDKs versions prior to 2.2.123 render Jinja2/Nunjucks prompt templates unsandboxed. This vulnerability allows for arbitrary code execution within the application process, impacting applications that format untrusted Jinja2 prompts. The issue is fixed in version 2.2.123, which now utilizes sandboxing for template rendering.
2026-11-11 2026mailcow CVE-2026-125757: Jinja footer SSTI executes in shared rspamd container news RCEWriteup on CVE-2026-125757 in mailcow: dockerized, detailing a server-side template injection vulnerability that allows domain administrators to execute commands within the shared rspamd container. Exploitation involves storing a malicious domain-wide footer. Affected versions prior to 2026-11 require an upgrade to 2026-11 or later to patch this critical vulnerability.
2026-11-11 2026mailcow: dockerized before 2026-11 server-side template injection (critical) news RCEWriteup of CVE-2026-125757 in mailcow:dockerized, a critical server-side template injection. This vulnerability allows domain administrators to execute arbitrary commands within the shared rspamd container by injecting a malicious Jinja macro into the domain-wide footer. Exploitation via the MOO_FOOTER prefilter enables reading or injecting mail for all tenants and disabling filtering. The affected versions include those prior to the 2026-11 release.
2026-11-11 2026CVE-2026-123198: unauthenticated Jinja SSTI in LMFDB render_knowl news Python RCELMFDB versions prior to commit 01829bd are vulnerable to CVE-2026-123198, an unauthenticated server-side template injection flaw in the `render_knowl` function. This vulnerability can be exploited by crafting specific query-string values. → insightcve.com
2026-11-11 2026LetMeOut: HICyberCon CTF 2026 Jinja2 sandbox escape writeup intermediate 2 min read PythonWriteup detailing the "LetMeOut" Jinja2 sandbox escape from HICyberCon CTF 2026. The exploit chains NFKC normalization vulnerabilities in `getitem` and `is_safe_attribute` to smuggle attribute names past guards. This allows access to `wrap_str_format`, which returns an unsandboxed function with `jinja2.sandbox` as its `__globals__`. This grants access to `pdb.breakpoint`, enabling arbitrary file reads and flag extraction, as demonstrated by `exploit.py`.
2026-11-10 2026SSTI lab walkthrough: email preview template injection intermediate 4 min readWalkthrough of Server-Side Template Injection (SSTI) in an email preview feature, demonstrating how template engines evaluating user input as expressions, rather than plain data, can expose server-side secrets. The lab details the attack chain: confirming expression evaluation via arithmetic, accessing server configuration through the template context like `config.flag`, and the root cause of users supplying the template. It maps to OWASP Top 10 Injection and suggests defenses such as separating template from data and using logic-less templating.
2026-11-10 2026HTB Takeoff: Go SSTI and n8n privilege escalation intermediate 1 min readWalkthrough of Hack The Box "Takeoff" details Go SSTI leading to RCE, environment variable disclosure via HMAC-SHA256, QR code forgery, ticket validation, n8n API exploitation, container mount misconfigurations, and Linux privilege escalation to root, covering reconnaissance with Rustscan and Nmap, initial foothold, enumeration, credential recovery, and flag capture.
2026-11-10 202630 minutes to root: AI found a Nunjucks SSTI RCE in local government software intermediate AI RCEAn AI agent, guided by a researcher, quickly discovered a severe vulnerability in OpenStad, a local government software. Within 30 minutes, the AI identified an authenticated Remote Code Execution (RCE) flaw related to Nunjucks Server-Side Template Injection (SSTI). This discovery highlights a significant risk for organizations using this software, as such vulnerabilities could allow attackers to gain full control of systems. The ease with which the AI found this exploit underscores the potential for automated tools to uncover critical security weaknesses in widely used software.
2026-11-10 2026From SSTI to a Django secret key (XBEN-056) intermediate 4 min read Python SecretsWriteup of XBEN-056 detailing a template injection vulnerability leading to the exposure of a Django secret key. The vulnerability occurred during a three-step registration process where a user-supplied name was compiled as template source, allowing double-brace expressions to access the framework's message store and signer. This enabled retrieval of the application's signing key material, potentially compromising cookies, reset links, and other signed values. Remediation involves treating user input as data, not template source, restricting exposed template context, and rotating compromised signing keys.
2026-11-09 2026Vulnfeed: 5 critical CVEs (LMFDB Jinja2 SSTI, OFBiz FreeMarker SSTI and more) news 1 min read RCEWriteup of five critical CVEs with CVSS scores of 9.0 or higher, including SuiteCRM workflow controller path traversal (CVSS 9.9), OpenHuman message injection (CVSS 9.8), SuiteCRM Studio Dropdown Editor HTML injection (CVSS 9.4), LMFDB Jinja2 server-side template injection (CVSS 9.3), and OpenHuman Yuanbao channel message processing (CVSS 9.0).
2026-11-09 2026openEQUELLA authenticated RCE via FreeMarker SSTI intermediate 2 min read RCEWriteup detailing authenticated RCE in openEQUELLA via FreeMarker SSTI. Versions prior to 2026.1.0 are vulnerable due to permissive FreeMarker defaults allowing arbitrary Java class instantiation via `?new()` and access to `?api`. Exploitation requires an authenticated non-guest user and the ability to store FreeMarker markup, leading to code execution on the application server. Patches in 2026.1.0 sandbox the configuration, blocking dangerous built-ins. Workarounds include limiting `CREATE_PORTLET` privileges.
2026-11-09 2026NocoDB server-side JavaScript execution via webhook template injection intermediate 1 min read RCEAdvisory detailing GHSA-2xx9-r3gm-mw87, a server-side JavaScript execution vulnerability in NocoDB. The vulnerability stems from Handlebars template injection via webhook configurations, where path-resolving helpers allow attackers to bypass restrictions and reach the JavaScript function constructor. This enables arbitrary server-side code execution and operating-system command injection with the server's privileges, impacting users with Creator role accounts.
2026-11-09 2026Craft CMS 5.10-5.11.2 remote code execution via template injection intermediate RCEA remote code execution vulnerability exists in Craft CMS versions 5.10.0 through 5.11.2. This flaw, stemming from template injection, allows an attacker to execute arbitrary code on the server. Users are strongly advised to update to a patched version immediately to mitigate this security risk. The specific patch is available in Craft CMS 5.11.3.
2026-11-09 2026Unauthenticated SSTI in LMFDB leads to RCE intermediate 2 min read Python RCEWriteup of unauthenticated Server-Side Template Injection (SSTI) in LMFDB, GHSA-6jw3-pmcx-2xww, leading to Remote Code Execution. The vulnerability arises from attacker-influenced data being passed unsanitized into `flask.render_template_string()`, particularly via the unauthenticated `/knowledge/render/<ID>` endpoint. This allows exploitation with a single HTTP GET request, enabling arbitrary Jinja syntax injection. The unsandboxed Jinja environment permits access to Python's object model, facilitating OS command execution via modules like `os`.
2026-11-08 2026Quest Foglight renderTemplateString SSTI RCE news RCEAdvisory for Quest Foglight for Databases renderTemplateString Server-Side Template Injection (SSTI) leading to Remote Code Execution (RCE). This vulnerability, documented as ZDI-CAN-29889, allows authenticated remote attackers to execute arbitrary code as SYSTEM. The flaw stems from improper validation of user-supplied strings within the renderTemplateString method, allowing malicious input to be processed by the template engine.
2026-11-08 2026Apache OFBiz MacroScreenRenderer FreeMarker SSTI to RCE (CVE-2026-82910) news RCELibrary for Apache OFBiz, addressing CVE-2026-82910, a critical Server-Side Template Injection (SSTI) vulnerability in MacroScreenRenderer. This flaw allows authenticated users with basic view permissions to execute arbitrary OS commands via a crafted GET request. The vulnerability stems from unchecked concatenation of FreeMarker expressions into the template engine, leading to unsandboxed code execution. Users should upgrade to version 24.09.08 to mitigate this risk.
2026-11-08 2026Elo Ruby backend RCE via unescaped interpolation in emitted templates news 1 min read RCEWriteup of GHSA-7w38-ggh5-v542 detailing a Ruby backend Remote Code Execution vulnerability in the Elo application security library. The flaw arises from unescaped '#' characters in emitted templates, specifically within string literals and labelled subtype constraints. This allows attacker-controlled Ruby code, including shell commands via backticks, to be executed at runtime when the Elo source is compiled to Ruby. The vulnerability bypasses Elo's cross-target portability guarantee, impacting Ruby compilation while leaving JavaScript, Python, and SQL targets benign.
2026-11-08 2026Elo JavaScript emitter sandbox escape via template evaluation news 2 min read RCELibrary emitting arbitrary code via template evaluation when handling attacker-controlled strings passed to `defaultFormats.elo.parse()`, the `bin/elo` CLI, or the live browser playground. The vulnerability arises from the `member_access` handler allowing unfiltered property names and the JavaScript emitter writing property names verbatim, enabling payloads like `({}).constructor.constructor('…')()` to execute as a live `Function` call.
2026-11-08 2026AWS Virtual Engineering Workbench publishing-component Jinja SSTI news 1 min read Python RCEAdvisory for GHSA-mf9w-9v72-cqm9 detailing a Jinja2 Server-Side Template Injection vulnerability in AWS Virtual Engineering Workbench's publishing component. Authenticated users with PRODUCT_CONTRIBUTOR roles can inject Jinja2 expressions into CloudFormation templates, allowing arbitrary code execution within the publishing Lambda function, with potential access to resources across all tenant projects. The vulnerability affects versions before commit c146486, which introduced template sandboxing and rejection of Jinja2 control sequences.
2026-11-07 2026SSTI2 ~ CyLab Security Academy Write-up | Server-Side Template Injection intermediate Bug BountyThis write-up by Affanhaxor details how to solve the SSTI2 challenge on the CyLab Security Academy. The challenge falls under Web Security and focuses on Server-Side Template Injection (SSTI). The author provides a step-by-step explanation of the process used to exploit the SSTI vulnerability within the challenge.
2026-11-07 2026ZDI-26-866 Quest Foglight renderTemplateString SSTI RCE intermediate RCEThis advisory details a Server-Side Template Injection (SSTI) vulnerability in Quest Foglight's `renderTemplateString` function, which can lead to Remote Code Execution (RCE). Attackers can exploit this flaw by crafting malicious input that, when processed by the `renderTemplateString` function, allows them to execute arbitrary code on the affected server. The specific technical details and impact of this vulnerability are detailed in ZDI-26-866. → zerodayinitiative.com
2026-11-07 2026LMFDB unauthenticated Jinja SSTI (CVE-2026-123198) intermediate Python RCELibrary advisory detailing CVE-2026-123198, an unauthenticated Jinja Server-Side Template Injection in LMFDB prior to commit 01829bd. The vulnerability in the `render_knowl` function allows attackers to inject Jinja expressions through request parameters or 'content' data to `/knowledge/render/<ID>`, enabling execution of arbitrary operating system commands as the web application user. → vulncheck.com

Frequently Asked Questions

What is server-side template injection?
SSTI happens when user input is concatenated into a template that the server then renders, instead of being passed as a parameter. Because templating engines support expressions and method calls, an attacker who controls part of the template can often execute arbitrary code in the language the engine runs on — Python, Java, PHP, JavaScript, and others have all seen real-world SSTI leading to remote code execution.
How do you test for SSTI?
The standard first probe is a math expression like {{7*7}} or ${7*7} — if the response contains 49 instead of the literal string, the input is being evaluated. From there, identify the engine through error messages or syntax fingerprinting and escalate to environment access, file reads, or code execution using engine-specific payloads. PortSwigger's SSTI research provides a complete methodology.
Which template engines are most affected by SSTI?
Jinja2 (Python), Twig (PHP), Freemarker and Velocity (Java), Smarty (PHP), and Mako (Python) all have well-documented SSTI exploitation paths. Even nominally sandboxed engines like Jinja2 have repeatedly had sandbox escapes published. Any engine that exposes object attributes, method calls, or filters is a candidate.

Weekly AppSec Digest

Get new resources delivered every Monday.