The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill includes a utility script scripts/update_schema.py that fetches the BIDS schema and Extension Proposals (BEPs) list from official sources. This is a legitimate maintenance function.
Evidence: Downloads from https://bids-specification.readthedocs.io/en/stable/schema.json and https://raw.githubusercontent.com/bids-standard/bids-website/main/data/beps/beps.yml.
[COMMAND_EXECUTION]: The skill provides instructions for installing and running standard neuroimaging tools and a local management script.
Evidence: Instructions to run python scripts/update_schema.py and CLI tools such as heudiconv, dcm2bids, and bids-validator.
[INDIRECT_PROMPT_INJECTION]: The skill processes external, researcher-provided datasets which constitutes a vulnerability surface for indirect prompt injection if those datasets contain malicious instructions in metadata fields.
Ingestion points: BIDSLayout (referenced in SKILL.md and core_workflows.md) indexes and reads metadata from JSON sidecars and TSV files (e.g., dataset_description.json, participants.tsv).
Boundary markers: Absent.
Capability inventory: The skill utilizes package installation tools (uv pip), local script execution, and shell-based data converters.
Sanitization: Absent; tools focus on schema validation rather than natural language instruction filtering.

bids