CVE-2026-43992: JunoClaw MCP write tools: raw BIP-39 mnemonic exposed as tool-call parameter
JunoClaw, an agentic AI platform built on Juno Network, had MCP write tools that accepted a raw BIP-39 mnemonic as an explicit tool-call parameter. This placed the wallet seed inside LLM tool-call JSON, where it could reach transport, logging or telemetry surfaces. The record states the issue is fixed in 0.x.y-security-1, and the CVSS 3.1 base score reported by GitHub is 9.8 (critical).
Overview
JunoClaw is an agentic AI platform built on Juno Network. According to the record, its MCP write tools, such as send_tokens, execute_contract, instantiate_contract, upload_wasm and ibc_transfer, accepted a mnemonic string as an explicit tool-call parameter. The weakness is classified as CWE-200, Exposure of Sensitive Information.
Impact
Because the BIP-39 seed was part of the LLM tool-call payload, it could be exposed to any transport, log or telemetry component between the LLM provider and the MCP process. The GitHub-assigned CVSS 3.1 base score is 9.8 with vector CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H, which indicates high confidentiality, integrity and availability impact as scored by the CNA.
Affected products and versions
- Dragonmonk111 Junoclaw: versions before v0.x.y-security-1.
The record does not list a specific set of fixed versions beyond the statement that the issue is fixed in 0.x.y-security-1.
Exploitation status
The KEV catalog does not list this CVE. No confirmed exploitation appears in the sources consulted. The SSVC exploitation field is recorded as none, and the EPSS probability is 0.00374.
Remediation and mitigation
- Upgrade JunoClaw to 0.x.y-security-1 or a later release that the maintainer publishes, as referenced in the official release page.
- Review the maintainer's GitHub security advisory for the full remediation guidance.
- Treat any wallet seed that may have passed through affected tool calls as potentially exposed, and plan rotation of funds according to your own risk assessment.
- Review logs, telemetry and proxy records from deployments that used the affected write tools for unexpected seed material.
Detection guidance
- Search logging and telemetry pipelines for mnemonic-like strings or tool-call arguments named mnemonic.
- Audit which MCP tools in your deployment accept secret material as normal parameters.
Limitations of the available information
The record does not describe the exact exploitation conditions, the full list of affected tool names beyond those listed, or whether any third-party component actually captured seed material. The fixes list is empty in the record, so the fixed version is taken from the description and the release reference.
Frequently asked questions
What is the core problem in CVE-2026-43992?
According to the record, MCP write tools in JunoClaw accepted a raw BIP-39 mnemonic as a tool-call parameter, placing the seed in LLM tool-call JSON.
Is this CVE known to be actively exploited?
No. The KEV catalog does not list it, and no confirmed exploitation appears in the sources consulted.
What should administrators do?
Upgrade to 0.x.y-security-1 or later, review logs and telemetry for seed material, and consider rotating funds based on your own risk assessment.
This analysis was drafted with AI assistance from verified source data and automatically checked against it.