CVE-2025-51482
published 2025-07-22CVE-2025-51482: Remote Code Execution in letta.server.rest_api.routers.v1.tools.run_tool_from_source in letta-ai Letta 0.7.12 allows remote attackers to execute arbitrary…
PriorityP180high8.8CVSS 3.1
AVNACLPRNUIRSUCHIHAH
ITWEXPLOITVulnCheck KEVInitial access
Exploited in the wild
EPSS
1.86%
76.6th percentile
Remote Code Execution in letta.server.rest_api.routers.v1.tools.run_tool_from_source in letta-ai Letta 0.7.12 allows remote attackers to execute arbitrary Python code and system commands via crafted payloads to the /v1/tools/run endpoint, bypassing intended sandbox restrictions.
Affected
1 ranges
| Vendor | Product | Version range | Fixed in |
|---|---|---|---|
| letta | letta | — | — |
Detection & IOCsextracted from sources · hover to see the quote
- →Monitor HTTP POST requests to /v1/tools/run containing a 'source_code' field with embedded Python import statements (e.g., 'import os', 'import subprocess') or OS command execution patterns, which indicate sandbox bypass attempts. ↗
- →Detect POST /v1/tools/run requests with Content-Type: application/json where the JSON body contains a 'source_code' key — this is the attack vector for CVE-2025-51482 RCE. ↗
- →Use FOFA query 'title="Letta"' to identify exposed Letta instances potentially vulnerable to CVE-2025-51482. ↗
- →A successful exploitation response returns HTTP 200 with Content-Type: application/json and a body containing the key 'tool_return' — use this to confirm RCE via the /v1/tools/run endpoint. ↗
- ·The vulnerability is unauthenticated (PR:N) — no credentials are required to reach the /v1/tools/run endpoint in the affected version, meaning any network-accessible Letta 0.7.12 instance is at risk. ↗
- ·The sandbox bypass is achieved by supplying arbitrary Python source code in the 'source_code' field of the JSON body — the intended sandbox restrictions in letta.server.rest_api.routers.v1.tools.run_tool_from_source are insufficient. ↗
CVSS provenance
nvdv3.18.8HIGHCVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H
vulncheck8.8HIGH
CVEs like this are exactly what “Exploited This Week” covers.
Every Monday: what got weaponized or added to CISA KEV in the last seven days — each CVE cross-linked to its PoC, Nuclei template, and detection rule. Free, one email a week, unsubscribe in one click.
GHSA
GHSA-2pgv-8585-494w: Remote Code Execution in letta
ghsa_unreviewed·2025-07-22
CVE-2025-51482 [HIGH] CWE-94 GHSA-2pgv-8585-494w: Remote Code Execution in letta
Remote Code Execution in letta.server.rest_api.routers.v1.tools.run_tool_from_source in letta-ai Letta 0.7.12 allows remote attackers to execute arbitrary Python code and system commands via crafted payloads to the /v1/tools/run endpoint, bypassing intended sandbox restrictions.
VulnCheck
letta letta Improper Control of Generation of Code ('Code Injection')
vulncheck·2025·CVSS 8.8
CVE-2025-51482 [HIGH] letta letta Improper Control of Generation of Code ('Code Injection')
letta letta Improper Control of Generation of Code ('Code Injection')
Remote Code Execution in letta.server.rest_api.routers.v1.tools.run_tool_from_source in letta-ai Letta 0.7.12 allows remote attackers to execute arbitrary Python code and system commands via crafted payloads to the /v1/tools/run endpoint, bypassing intended sandbox restrictions.
Affected: letta letta
Required Action: Apply remediations or mitigations per vendor instructions or discontinue use of the product if remediation or mitigations are unavailable.
Exploitation References: https://dashboard.shadowserver.org/statistics/honeypot/vulnerability/map/?day=2025-10-24&host_type=src&vulnerability=cve-2025-51482; https://app.crowdsec.net/cti/cve-explorer/CVE-2025-51482; https://dashboard.shadowserver.org/statistics/honeyp
No detection rules found.
Nuclei
Letta Letta 0.7.12 - Remote Code Execution
nuclei·CVSS 8.8
CVE-2025-51482 [HIGH] Letta Letta 0.7.12 - Remote Code Execution
Letta Letta 0.7.12 - Remote Code Execution
Letta 0.7.12 is vulnerable to remote code execution via POST /v1/tools/run in letta.server.rest_api.routers.v1.tools.run_tool_from_source, allowing attackers to execute arbitrary Python and OS commands via crafted tool source code.
Template:
id: CVE-2025-51482
info:
name: Letta Letta 0.7.12 - Remote Code Execution
author: RaghavArora14
severity: high
description: |
Letta 0.7.12 is vulnerable to remote code execution via POST /v1/tools/run in letta.server.rest_api.routers.v1.tools.run_tool_from_source, allowing attackers to execute arbitrary Python and OS commands via crafted tool source code.
impact: |
Unauthenticated attackers can execute arbitrary Python code through crafted tool source code in the /v1/tools/run endpoint, achieving remote co
Greynoiseio
NoiseLetter August 2025
blogs_greynoiseio
NoiseLetter August 2025
CVE Disclosure Early Warning Get an early warning when traffic spikes indicate a high likelihood of new disclosures
Compromised Asset Detection Find out immediately if an asset communicates with a malicious IP address
Vulnerability Prioritization Get real-time insight into active exploitation trends to better understand risk and severity
SOC Efficiency Filter out noisy, low priority and false-positive alerts from mass internet scanners
Incident Investigation Add context to incidents to speed the determinations of scope and timelines
Threat Hunting Quickly identify anomalous behavior and enrich your threat hunting campaigns
Why GreyNoise
CVE Disclosure Early Warning Get an early warning when traffic spikes indicate a high likelihood of new disclosures
Compromised Asset Detection Fin
arXiv
AXE: An Agentic eXploit Engine for Confirming Zero-Day Vulnerability Reports
arxiv_fulltext·2026-02-15
AXE: An Agentic eXploit Engine for Confirming Zero-Day Vulnerability Reports
AXE: An Agentic eXploit Engine for Confirming Zero-Day Vulnerability Reports
Amirali [email protected]
Drexel University, Philadelphia, PA
Tu [email protected]
Drexel University, Philadelphia, PA
Kostadin [email protected]
Virginia Commonwealth University, Richmond, VA
Preetha [email protected]
Drexel University, Philadelphia, PA
## Abstract
Vulnerability detection tools are widely adopted in software projects, yet they often overwhelm maintainers with false positives and non-actionable reports. Automated exploitation systems can help validate these reports; however, existing approaches typically operate in isolation from detection pipelines, failing to leverage readily available metadata such as vulnerability type and source-code
2025-07-22
Published
Exploited in the wild