neurokit
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill package includes a file neurokit2_audit_result_v1.json that asserts the skill has passed security audits (e.g., 'security': 'PASS', 'deployable': true). This is a self-referential injection designed to bias security assessments.
- Ingestion points: The audit metadata in neurokit2_audit_result_v1.json is processed as a skill resource.
- Boundary markers: The file lacks any delimiters or instructions for an agent to treat the safety claims as untrusted content.
- Capability inventory: Analysis of the skill's scripts and referenced docs reveals no dangerous shell commands, unauthorized file system access, or network exfiltration beyond its stated data processing purpose.
- Sanitization: No sanitization or validation of the self-referential metadata is performed.
- [EXTERNAL_DOWNLOADS]: Fetches and installs the development version of the NeuroKit library from the project's official GitHub repository.
- Evidence: uv pip install https://github.com/neuropsychology/NeuroKit/zipball/dev in SKILL.md.
Audit Metadata