Skip to content
Skills Library
Science & research · Planning

md-study

/md-study

Study-level planning and workflow routing for MDClaw. Use for scientific questions, comparative or campaign studies, plan-only requests, and requests to carry planned MD jobs through analysis to an evidence-backed answer.

Repository
mdclaw ↗
Licence
MIT
Last repository change
26.10.2026

Assessment and security

90/ 100Class: good
Format25 / 25
Frontmatter with name and description, valid name, description up to 1,024 characters, name matching the folder.
Discoverability25 / 25
Description explains when to use the skill, has a reasonable length and uses third person.
Content25 / 35
At least 150 words and at most 500 lines, headings, step lists, examples or code.
Hygiene15 / 15
No placeholders such as TODO, no Windows paths, with a licence.
Riskno signals detected in the scan
Files in this package
7
Scripts
0
HTML comments
0
URLs in SKILL.md
0
Deprecation

No signals detected. Repository changed on 26.10.2026, without references to retired models or APIs.

Licence

MIT, open: use and modify while retaining author attribution.

What it does and when to use it

Study-level planning and workflow routing for MDClaw. Use for scientific questions, comparative or campaign studies, plan-only requests, and requests to carry planned MD jobs through analysis to an evidence-backed answer.

Original author description

Study-level planning and workflow routing for MDClaw. Use for scientific questions, comparative or campaign studies, plan-only requests, and requests to carry planned MD jobs through analysis to an evidence-backed answer.

How it works

  1. When To Use This Skill
  2. Step 0: Parse and Confirm
  3. Workflow
  4. Guardrails
Original SKILL.md headings · 82 lines, 606 words
SKILL.md preview
---
name: md-study
description: "Study-level planning and workflow routing for MDClaw. Use for scientific questions, comparative or campaign studies, plan-only requests, and requests to carry planned MD jobs through analysis to an evidence-backed answer."
---

# MD Study

You are a computational biophysics expert helping users turn scientific
questions into MDClaw studies.

Follow `skills/common/preamble.md`, `skills/common/run-loop.md` (the canonical
node loop), and `skills/common/tool-output.md` for error handling. The workflow
below routes to the focused planning pages.

## When To Use This Skill

Use this skill when the user asks a scientific or campaign-level question:
comparing WT vs mutant, apo vs holo, temperatures, force fields, or candidates;
asking which simulations answer a biological/physical question; or asking for
production length, replicates, observables, controls, or decision criteria.

All MD workflows use a study plan and the canonical `study_dir/jobs/<job_id>`
layout. If the user gives a concrete target and just asks to run it, do **not**
do campaign design — follow `skills/md-study/direct-run.md` instead.

## Step 0: Parse and Confirm

Extract parameters from the user's request and present a summary.

| Parameter | Value |
|-----------|-------|
| Scientific question | (one sentence, copied or restated from the user) |
| Study directory | (path, e.g. `studies/<study_id>`) |
| Stop after | plan / preparation / equilibration / production / analysis / scientific answer (from the current request; do not persist) |
| Interaction mode (`execution_mode`) | `autonomous` (default) / `human_in_the_loop` |
| Variants / planned jobs | (WT vs mutant, apo vs holo, etc., if named) |
| Solvent regime | `explicit` (default) / `implicit` / `vacuum` / `membrane` |
| Compute budget | (free text, e.g. "1x A100 for 7 days"; or "not specified") |
| Other | (only parameters the user explicitly named) |

Pick `autonomous` unless the user explicitly asks for checkpoint-by-checkpoint
confirmation. This mode controls pauses within the requested work; it does not
decide how far to run. Propagate it to each registered job's `progress.json` so
downstream skills inherit the interaction policy. Determine the stopping point
from `skills/common/run-loop.md`; full handoff behavior is in
`skills/md-study/handoff-routing.md`.

## Workflow

1. Parse the request; set `execution_mode` per Step 0 (default `autonomous`).
2. If this is one concrete target, follow `skills/md-study/direct-run.md` and
   skip campaign design. Otherwise continue with campaign planning.
3. Ground the design in databases and literature per
   `skills/md-study/literature-lookup.md`. Do not pick structures, comparison
   cells, or observables from training-data memory.
4. Restate the question in one sentence, then translate it into an MD goal
   (what structural/dynamical/interaction behavior MD can measure).
5. Choose the study-level `solvent_regime` and design the smallest job set that
   answers the question (prefer one baseline/control + one test variant), a
   short analysis list, and support/against/inconclusive decision criteria. See
   `skills/md-study/minimal-plan-schema.md`.
6. Compute budget: follow `skills/md-study/compute-budget.md` when the user or
   harness specified compute. Also follow it for a scientific-answer request
   that omitted production length, using its labeled default assumption so
   production length is determined before handoff. For other requests, omit
   the `budget` block when compute was not mentioned.
7. Record the plan and register jobs per `skills/md-study/register-jobs.md`.
8. Hand off only as far as the current request requires, following
   `skills/md-study/handoff-routing.md`.

## Guardrails

- Do not select starting structures, comparison cells, or analysis observables
  purely from training-data memory; use the lookups in
  `skills/md-study/literature-lookup.md` and record consulted PDB IDs / PMIDs
  under `notes.references`.
- Do not treat visual QA or simple RMSD plots as scientific validation.
- Do not make the plan so detailed that later agents must satisfy fragile fields
  before running ordinary MD, and do not block downstream execution on an
  incomplete plan field — ask only when a missing value is needed for a safe
  next action.
- Keep execution state in each job DAG; the study plan is intent, not a
  replacement for node artifacts or `progress.json`.
- Do not infer permission to execute a stage from plan `workflow_steps`, a
  stored `execution_mode`, or an earlier HPC submission.
Files in this package (7)
  • SKILL.md 4 KB
  • compute-budget.md 4 KB
  • direct-run.md 1 KB
  • handoff-routing.md 2 KB
  • literature-lookup.md 2 KB
  • minimal-plan-schema.md 2 KB
  • register-jobs.md 2 KB

Versions

The database has no other skills named md-study.

Database snapshot: 13.11.2026 · package version 0dbd01fae039 (SHA-256 digest of the entire package)

Downloads in the last 30 days: 0 · ranking

Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.

Similar skills

  • 90research-start
    Start a research project, investigate a new research idea, or decide where to begin a data-science/ML study. Use for requests such as start research, explore an idea, plan a study, or define the research pipeline. Route clarification, component structure and planning before implementation.
  • 100experiment-design
    Design experiments and studies from research questions. Triggers when user says: 'design experiment', 'study design', 'experimental setup', 'how should I test this', 'plan my study', 'ablation study', 'baseline comparison', 'research protocol', 'pilot study', 'sample size', 'how…
  • 100scientific-writing-draft
    Use when a researcher requests new scientific manuscript prose from supplied study records, results, notes or identified literature.
  • 90research-interview-synthesis
    Synthesize multiple user interviews into evidence-backed themes, insights, limitations, and next research actions. Use for: interview synthesis, customer research findings, usability study themes, discovery interview analysis.
  • 100study-design
    Design a research study. Use when the user says "study design", "design my research", "what research method should I use", "how do I test this hypothesis", "research plan", "methodology section", "experimental design", "survey design", "qualitative study design", "how to…
  • 100packmol
    Use when writing, reviewing, debugging, or explaining packmol input files and packing strategies for molecular dynamics simulations.

Feedback and reports

Your message goes to Bartosz Sokoliński. An email address is needed only for a reply. See the privacy policy.