Highflame ramparts versions 0.8.1 through 0.8.10 contain a scan bypass that permits malicious payloads to be hidden in nested compiled Python files (.pyc) inside __pycache__ directories. The bundle validation logic, collect_bundle_siblings and walk_bundle_subdir, ignores nested directories and .pyc files, allowing the bundle to pass scanning. When the skill runs, CPython automatically executes the compiled file, giving an attacker the ability to run arbitrary code on the host, thereby compromising the system's integrity and confidentiality.

The vulnerability affects the highflame‑ai Ramparts product in all releases from 0.8.1 to 0.8.10.

With a CVSS score of 7.1 the risk is moderate to high. Exploitation requires an attacker to supply a skill bundle with a nested __pycache__ containing an unchecked .pyc file, a scenario possible when skill bundles are imported from external sources with insufficient validation. Although EPSS data is unavailable and the vulnerability is not listed in CISA KEV, the ability to execute hidden payloads poses a serious threat if the host accepts untrusted bundles.