Skip to content

ddonnyy/idus

v0.8.0Apache-2.0

IDUs — IDU-scientist: academic drafting, manuscript review and expert-informed community rules.

IDUs — IDU-scientist

Agent Skills for English scientific writing, manuscript review and expert-derived community rules.

npx skills add DDonnyy/IDUs --skill idus-draft --skill idus-review --skill idus-configure --copy

Install only the skills you need. Each is self-contained; Python is optional for review/configuration helpers and unnecessary for draft. See installation for copied skills/plugins.

SkillRequestOutput
idus-draftFirst version from research notes, findings or rough proseEnglish draft, specific gap markers and short author notes
idus-reviewAudit or authorised edits of existing proseLocated findings, evidence/wording verdicts and a diff where applicable
idus-configureProject settings or reusable guidance from expert editsScoped rules/terms or a sanitised contribution proposal

For example: “Use idus-draft. Draft Methods from these notes; keep missing details as TODOs.” For direct use without installation, copy the standalone prompt, or use the draft prompt / configure prompt.

The author supplies research and intended claims. Preserve uncertainty, defined terms, numbers and conditions; missing support stays explicit. Clear wording can remain unchanged. Style preferences are scoped, and an expert edit is evidence to assess rather than a universal rule.

Project settings live in .idus/config.json in the manuscript workspace. Shared-rule preparation checks semantic overlap, provenance and fix/keep examples; maintainers decide acceptance. See settings, contributions and source provenance.

Contribute and develop

Edit canonical sources, not generated skills/, then run:

python scripts/sync_distribution.py --write
python -m unittest discover -s scripts/tests -v
python scripts/check_repository.py
python scripts/build_release.py

Python 3.10+, standard library. CONTRIBUTING.md covers PRs; maintenance covers CI, releases and GitHub settings. Tests check helper/packaging behaviour, not scientific validity or model reliability.

Apache-2.0 for original toolkit contributions: LICENSE, NOTICE, third-party rights. Private reference archives are not distributed.