PraxIS — IB Sciences Practical & IA Suite
Curriculum-layer Agent Skills for the IB DP Group 4 sciences — Biology, Chemistry, and Physics — built from the official IB sciences guides, teacher support material, sciences experimentation guidelines, the official assessed samples, and a large corpus of real graded IAs used for calibration.
Status: in development (v0.3.0). Skills are drafted and being tested live. Not yet published.
What this is
ib-sciences-praxis-coach powers the whole practical-science workflow for a DP sciences teacher. The four-criteria IA model (Research design / Data analysis / Conclusion / Evaluation, 24 marks) is shared across the sciences; the coach applies subject-specific conventions (chemistry uncertainties and titration protocol, physics error propagation and linearization, biology's qualitative judgment) once the subject is named:
- Design guidance — co-shape a student's IA idea into a unique, quantifiable research question and a feasible, criteria-ready methodology (scaffolding, never ghost-writing)
- Review — criterion-referenced critique and comments on a student's plan or draft
- Marking support — support the teacher in marking a final report against the official criteria with evidence-based justification, calibrated against 849 recorded graded IAs (596 with full four-criteria scores: 220 Biology, 199 Chemistry, 177 Physics)
- Lesson practical design — plan practicals and activities for lessons from a topic and the equipment the school actually has
- Submission pack — the completeness checks and files needed for IB submission (word count, appendices rule, cover-sheet fields, teacher's per-criterion comments, moderation-sample advice)
You provide what you have or what you want to accomplish — no database required, and nothing is invented: equipment, reagents, and quantities always come from you or are flagged "determine in trial run". Optional companion skill: equipment-inventory maintains a private YAML lab inventory the coach can read automatically.
Install
Hermes Agent
hermes skills tap add LittleJakub/PraxIS
Manual (same effect, no tap index): clone into the profile skills directory —
git clone https://github.com/LittleJakub/PraxIS.git "$LOCALAPPDATA/hermes/skills/ib-sciences-praxis"
(The repo is also a portable Agent Plugins package — plugin.json + skills/ — so plugin-capable hosts can load the whole thing as one plugin.)
OpenClaw (and other Agent Skills hosts)
The suite is a standard Agent Skills repo (SKILL.md with frontmatter per agentskills.io), so any host that supports the open standard can load it:
# OpenClaw / other standard hosts: clone into the skills directory
git clone https://github.com/LittleJakub/PraxIS.git <your-skills-dir>/ib-sciences-praxis
# Or via the common standards-based installer
npx skills add LittleJakub/PraxIS
# Or via GitHub CLI (v2.90+), with provenance
gh skill install LittleJakub/PraxIS --agent <host>
If your host is not listed, check its current docs for where skills live (Claude Code, Codex, Gemini CLI, Cursor, OpenClaw, and others all read the standard SKILL.md format).
Optional companion: scientific-computing layer
PraxIS is the subject-domain layer; for heavy data analysis (statistics, plotting, literature retrieval) install the MIT-licensed K-Dense scientific-agent-skills collection alongside it — see DEPENDENCIES.md for the dependency and sync policy.
How it works
Ask your agent, for example:
- "My student wants to investigate enzymes for their Biology IA. Help me guide their design — SL, 10 hours, we have colorimeters and water baths." → Mode A
- "Here is my chemistry student's draft — review it against the criteria." → Mode B
- "Help me with the marking of this physics report — moderation comments please." → Mode C
- "Design a 50-minute practical for 'gas exchange in plants' with what we have." → Mode D
- "What do I need to submit for the moderation sample this session?" → Mode E
The skill encodes the criteria, assessment methodology, data rules, safety/ethics, subject conventions, and examiner patterns from the official samples plus real-IAs calibration — all distilled into references/ in the author's own words.
What's deliberately NOT here
- Generic scientific tooling — install the K-Dense scientific-agent-skills collection separately (see above).
- IB-copyrighted content — criteria and guidance are paraphrased in the author's own words; the official IB publications remain the authority. The skill can consult the teacher's own copies (e.g.
IB_DOCS_DIR) for exact descriptor wording. - School-specific data — real inventories, school policies, and student work stay in private files outside this repo.
- Anything a student could submit as their own IA — the coach scaffolds and evaluates; it never writes student work.
License & disclaimer
MIT. Teacher-authored; not affiliated with or endorsed by the International Baccalaureate Organization. Marking and feedback are starting points for qualified teacher judgment — verify against the official guide, school policy, and your own professional standards. All practicals require the teacher's own risk assessment and school approval.