nwb-conversion
Fail
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the user to install several Python dependencies, including a version of the 'tifffile' package that is highly suspicious.
- Evidence: The compatibility section and installation instructions in
SKILL.mdspecifytifffile==2026.9.20. - Analysis: The version
2026.9.20is dated in the future relative to current legitimate releases, which is a characteristic pattern of malicious package uploads to public registries intended to facilitate supply chain attacks. - [REMOTE_CODE_EXECUTION]: By requiring the installation of a suspicious package version from an external registry, the skill facilitates the execution of potentially malicious code during the package installation process.
- Evidence: The
uv pip install ... tifffile==2026.9.20command provided inSKILL.md. - Analysis: Package installation can execute arbitrary code contained in the package's setup or initialization scripts. The use of a non-standard, future-dated version string increases the risk that the source of the package is attacker-controlled.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external TIFF and CSV files, creating a vulnerability surface where instructions embedded in data could influence downstream agent behavior.
- Ingestion points:
scripts/convert_session.pyreads data from local paths specified in a configuration JSON for the TIFF image, frame timestamps CSV, and position CSV. - Boundary markers: The skill does not implement specific delimiters or instructions for the agent to ignore potentially malicious content embedded in the acquisition data or metadata.
- Capability inventory: The conversion script has the capability to read and write files on the local system. It does not possess network access or shell execution capabilities that could be directly exploited via data injection.
- Sanitization: The script implements several data validation steps, including numeric type conversion, finite value checks, and timestamp monotonicity verification. It also leverages
pynwb.validateandnwbinspectorto ensure the resulting file adheres to the NWB schema.
Recommendations
- AI detected serious security threats
Audit Metadata