The JunoClaw agentic AI platform’s upload_wasm service accepted a filesystem path supplied by an agent without validating the location, symlink target, size, or format of the file. This allowed the agent to instruct the system to upload the contents of any file that the agent could access, effectively exposing arbitrary files and enabling the upload of malicious WebAssembly binaries. The flaw is a classic path‑traversal and input‑validation issue, corresponding to CWE‑20, CWE‑22, CWE‑59 and CWE‑73. The primary impact is that an attacker with control over the agent can read, upload, or replace arbitrary files on the host, potentially leading to privilege escalation or remote code execution if malicious wasm is served.

All releases of the Dragonmonk111 JunoClaw platform prior to the 0.x.y-security-1 update are affected. No specific sub‑versions are listed as affected, but any build that has not been updated to this security patch is vulnerable.

The CVSS score of 8.5 indicates high severity. The EPSS score is not available, so the current probability of exploitation is unknown, but the vulnerability is likely exploitable over the network by malicious agents. It is not listed in the CISA KEV catalog. An attacker would need control of an agent to supply a malicious file path; once supplied, the system would blindly upload that data, allowing the attacker to exfiltrate or inject files.