Threat Intelligence Database

CVE-2026-62676 is a vulnerability in omnigent-ai's omnigent v0.1.0 and earlier versions before 0.3.0. The issue lies in the shared shell-command parser which fails open, allowing unrecognized commands to bypass critical policy checks. This can enable an agent intended to be confined to push code or secrets to attacker-controlled repositories, push to protected branches, or escape workspace confinement. The vulnerability stems from the parser returning 'None' (interpreted as allow) for unrecognized commands, rather than denying them.

CVE-2026-62674 is a critical code injection vulnerability in the omnigent-ai omnigent framework prior to version 0.3.0. It allows an authenticated user with edit access to a session to replace a shared agent bundle lacking a session ID and inject attacker-controlled commands. These commands execute with the permissions of the Omnigent runner process, potentially exposing sensitive files, credentials, workspace data, internal services, and affecting runner availability. The issue is fixed in version 0.3.0.

Omnigent is an open-source AI agent framework and meta-harness for orchestrating coding agents. Prior to 0.3.0, an authenticated user can upload a session-scoped agent bundle with an absolute or traversal-containing os_env.cwd value because omnigent/spec/parser.py stores the value verbatim and omnigent/spec/validator.py does not constrain it. On a runner where OMNIGENT_RUNNER_WORKSPACE is unset, omnigent/runner/resource_registry.py preserves the attacker-controlled path and omnigent/inner/os_env.py uses the resolved path as the environment root and copytree source. The _assert_within_cwd check then treats that attacker-selected root as trusted, allowing sys_os_read, write, edit, and shell tools to access runner files and environment secrets outside the intended workspace. This issue is fixed in version 0.3.0.