The Agent Skills Directory

[SAFE]: The skill performs dataset profiling and auditing as described. No malicious behavior was detected.\n- [PROMPT_INJECTION]: The instructions in SKILL.md are focused on dataset analysis and do not contain attempts to override agent behavior or safety filters. The skill also has a data ingestion surface through NIfTI files (scripts/profile_imaging_dataset.py), but these are binary formats not prone to prompt injection. Boundary markers and sanitization are not applicable for binary imaging data.\n- [DATA_EXFILTRATION]: No hardcoded credentials, sensitive path access, or network calls were identified. The code processes medical imaging data locally.\n- [REMOTE_CODE_EXECUTION]: The skill uses standard Python libraries (numpy, nibabel) and does not download or execute remote code.

profile-imaging