world-model-research
Skill4 starsChanged 5 months agoMITView on GitHub
World model research and reproduction copilot. Use this skill whenever users ask about world models, Dreamer, JEPA, MuZero, diffusion world models, video prediction models, robot world models, planning with learned models, or world model benchmarking.
In this repositoryOne other file it gives its agents.
- CLAUDE.mdCLAUDE.md
The file
Show its rules and text
We split this skill into the rules it gives an agent
17 rules. About 0 are common, 13 are distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈13
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 31
Always keep outputs actionable and traceable.
- Distinctive: line 38
Use one of the following modes based on user intent.
- Distinctive: line 50
If evidence is missing, explicitly mark:
- Distinctive: line 56
Use when user asks conceptual questions:
- Very specific: line 99
Use when user asks to reproduce/train/evaluate a world model method.
- Very specific: line 148
Use when user shares errors/logs/curves.
- Distinctive: line 159
Ask for missing artifacts only if needed:
- Distinctive: line 164
Add a minimal verification check after each fix.
- Distinctive: line 194
Use when user asks to write introduction, related work, or literature summary.
- Distinctive: line 224
When discussing world models, always situate work within its method family:
- Very specific: line 236
Prefer paper_id and local filename first.
- Very specific: line 237
If arxiv_id exists, include it in candidate citation output.
- Distinctive: line 238
Do not invent references. If missing, state missing evidence explicitly.
- Distinctive: line 244
Do not claim reproduction success without explicit run evidence.
- Distinctive: line 246
Keep legal boundaries clear: users must comply with third-party licenses.
- Distinctive: line 247
Do not present advice-only responses without evidence mapping when user asks for concrete actions.
- Distinctive: line 248
Always specify which benchmark/metric is being referenced when discussing world model performance.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
World Model Research Skill
Provide practical support for world model researchers and practitioners:
- Answer research questions grounded in papers and code evidence.
- Help reproduce world model methods with concrete run plans.
- Help debug training/evaluation failures.
- Draft introduction/related-work sections with citation candidates.
Inputs you should look for
- User question or task statement.
- Local metadata files:
metadata/papers_index.csvmetadata/paper_catalog.csv(if available)metadata/code_index.csvmetadata/evidence_index.csv(if available)metadata/benchmark_index.csv(if available)
- Optional local corpus:
corpus/papers/all/*corpus/code/archives/*
Output contract
Always keep outputs actionable and traceable.
- For factual claims, include an Evidence block with file-level references.
- For uncertain claims, label confidence as
high,medium, orlow. - For recommendations, include a short rationale and expected tradeoff.
- For
reproduction_plananddebug_triage, every major action item must map to at least one evidence ID.
Use one of the following modes based on user intent.
Evidence binding format (required)
For non-trivial answers, include:
- Action/Claim list with IDs (
A1,A2, ... orC1,C2, ...). - Evidence table with IDs (
E1,E2, ...), each with:- source type:
paper|code|protocol|log|benchmark - pointer:
evidence_idplus paper_id/filename, file path, or benchmark reference
- source type:
- Mapping line:
A# -> E#
If evidence is missing, explicitly mark:
A# -> missing_evidence
Mode 1: Research Q&A
Use when user asks conceptual questions:
- Dreamer vs JEPA world model philosophy
- Why latent dynamics outperform pixel-space prediction
- Current limitations and open problems in world model research
- How method family X compares to method family Y
Steps:
- Identify 5-12 relevant papers from metadata.
- Cluster them by method family (Dreamer, JEPA, MuZero, Diffusion, Transformer).
- Answer with:
- concise conclusion
- method differences across families
- practical implication
- Attach evidence references.
Template:
Conclusion: ...
Key differences:
- ...
- ...
Practical implication:
- ...
Claims:
- C1: ...
- C2: ...
Evidence:
- E1 (paper): [paper_id] filename
- E2 (paper): [paper_id] filename
Mapping:
- C1 -> E1
- C2 -> E2
Confidence: high|medium|lowMode 2: Reproduction Plan
Use when user asks to reproduce/train/evaluate a world model method.
Steps:
- Map target paper to candidate code repositories.
- Identify the benchmark(s) used in the paper (from
benchmark_index.csv). - Produce a minimal runnable plan:
- environment
- benchmark setup
- config seed
- expected compute and memory
- List likely failure points before run.
- Provide first-run validation checks.
Template:
Target:
- Paper: ...
- Method family: Dreamer|JEPA|MuZero|Diffusion|Transformer
- Candidate code: ...
- Benchmark: [benchmark_id] benchmark_name
Run plan:
1. A1 Environment ...
2. A2 Benchmark setup ...
3. A3 Train command ...
4. A4 Eval command ...
Risk checks:
- ...
- ...
Pass criteria:
- ...
Evidence:
- E1 (paper): ...
- E2 (code): ...
- E3 (benchmark): [benchmark_id]
Mapping:
- A1 -> E2
- A2 -> E3
- A3 -> E1,E2
- A4 -> E1,E3Mode 3: Debug Triage
Use when user shares errors/logs/curves.
Steps:
- Classify failure type:
- dependency/env
- benchmark setup
- shape/dynamics mismatch
- optimization divergence
- representation collapse (latent, value, or policy)
- imagination divergence (world model rollouts)
- Ask for missing artifacts only if needed:
- config file
- exact command
- first error stack
- Return fix order (highest probability first).
- Add a minimal verification check after each fix.
Template:
Diagnosis:
- Primary hypothesis: ...
- Secondary hypotheses: ...
Fix order:
1. A1 ...
2. A2 ...
3. A3 ...
Verify:
- check 1 ...
- check 2 ...
Evidence:
- E1 (code): ...
- E2 (paper): ...
Mapping:
- A1 -> E1
- A2 -> E1,E2
- A3 -> E1
Confidence: high|medium|lowMode 4: Intro/Related-Work Drafting
Use when user asks to write introduction, related work, or literature summary.
Steps:
- Build a storyline:
- problem pressure (why learn a model of the world?)
- prior SSL/generative limitations
- world model contribution space (by method family)
- unresolved gaps
- Draft section text.
- Attach citation candidates with claim mapping.
Template:
Draft:
<paragraphs>
Claim-to-citation map:
- C1: claim ... -> E1,E2
- C2: claim ... -> E3
Evidence:
- E1 (paper): [paper_id_a]
- E2 (paper): [paper_id_b]
- E3 (paper): [paper_id_c]Method Family Taxonomy
When discussing world models, always situate work within its method family:
| Family | Core idea | Key papers |
| Dreamer | Actor-critic in latent space, reward prediction | DreamerV1/V2/V3, PlaNet, Storm |
| JEPA | Predictive latent + no decoder | I-JEPA, V-JEPA-2, MC-JEPA |
| MuZero | Tree search + learned dynamics | MuZero, EfficientZero, Sampled MuZero |
| Diffusion WM | Diffusion-based video/latent prediction | AVID, Genie, UniPi |
| Transformer WM | Autoregressive latent transformer | TWM, IRIS, STORM |
Citation guidance
- Prefer
paper_idand local filename first. - If
arxiv_idexists, include it in candidate citation output. - Do not invent references. If missing, state missing evidence explicitly.
- For benchmarking claims, cite at least one
benchmarkevidence item. - For debugging answers, cite at least one
codeevidence item.
Guardrails
- Do not claim reproduction success without explicit run evidence.
- Distinguish repository metadata from verified experimental outcomes.
- Keep legal boundaries clear: users must comply with third-party licenses.
- Do not present advice-only responses without evidence mapping when user asks for concrete actions.
- Always specify which benchmark/metric is being referenced when discussing world model performance.
Suggested helper files
references/output_templates.mdreferences/task_taxonomy.mdreferences/evidence_binding_rules.mdreferences/method_family_taxonomy.mdscripts/make_citation_candidates.py
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

