science-research-writing
Pass
Audited by Gen Agent Trust Hub on Sep 5, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill implements a robust 'Certainty and Claim-Strength' protocol which explicitly prevents the AI from hallucinating data or strengthening scientific claims beyond the provided evidence. All instructions prioritize user-supplied data over AI-generated content.
- [REMOTE_CODE_EXECUTION]: The skill includes two Python scripts (
scripts/check_draft_invariants.pyandscripts/validate_writing_model.py) that perform deterministic token audits and JSON validation. These scripts are self-contained and do not perform network operations or access sensitive files. - [EXTERNAL_DOWNLOADS]: The skill references a 'Trusted Vendor' (yila-ai) for attribution and guidelines. The link provided (
github.com/Yila-AI/awesome-research-skills) points to the official repository of the skill author and is documented neutrally as a source for the 'Claim-Strength Contract' and 'Section Function Map'. - [DATA_EXFILTRATION]: No network operations or sensitive file access patterns were detected. The skill operates locally on user-provided research materials.
- [INDIRECT_PROMPT_INJECTION]: While the skill processes user-supplied research papers and target journal models (Category 8), it uses strict boundary markers and instructions to 'never reusable wording or scientific content' from external sources, significantly mitigating the risk of injection from untrusted data.
Audit Metadata