bids
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill includes a maintenance script (
scripts/update_schema.py) that fetches schema definitions and project metadata from official BIDS documentation sites (bids-specification.readthedocs.io) and the project's GitHub organization. These are well-known, trusted sources for the neuroscience community. - [COMMAND_EXECUTION]: The skill facilitates the use of community-standard command-line interfaces such as
heudiconv,dcm2bids, andbids-validatorfor converting DICOM files to BIDS and verifying dataset compliance. These operations are consistent with the skill's stated purpose. - [INDIRECT_PROMPT_INJECTION]: The skill operates on neuroimaging datasets and associated metadata which are sourced from external files, presenting a theoretical injection surface.
- Ingestion points: Local BIDS dataset paths, JSON metadata sidecar contents, and DICOM headers processed by the agent or referenced tools (e.g.,
pybids,heudiconv). - Boundary markers: No explicit delimiters or instructional guardrails are used to separate potentially untrusted data content from agent instructions in the provided documentation.
- Capability inventory: The skill possesses capabilities for reading/writing local files and executing shell commands (via installation and conversion workflows).
- Sanitization: The skill relies on the internal parsing and validation logic of established third-party neuroimaging libraries to handle data safely.
Audit Metadata