Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add rjmurillo/ai-agents --skill world-model-diagnosticgit clone --depth 1 https://github.com/rjmurillo/ai-agentsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/rjmurillo/ai-agents/world-model-diagnostic)<a href="https://agentmods.dev/skills/rjmurillo/ai-agents/world-model-diagnostic"><img src="https://agentmods.dev/badge/skills/rjmurillo/ai-agents/world-model-diagnostic/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/rjmurillo/ai-agents/world-model-diagnostic"><img src="https://agentmods.dev/badge/skills/rjmurillo/ai-agents/world-model-diagnostic.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 2 findings, up to high
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- high Data Exfiltration · line 291 Code or instructions that leak agent conversation context to external services, potentially exposing sensitive user interactions.Fix: Remove any code that sends prompts, responses, or session data externally. Preserve user privacy; never exfiltrate conversation content.
- high Prompt Injection · line 304 Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.Fix: Audit all comments and invisible characters. Remove any instructions that direct the agent to perform unauthorized actions. Use plain, reviewable content.
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00094 | $0.03216 |
| Opus 5 | $0.00047 | $0.01608 |
| Sonnet 5 | $0.00019 | $0.00643 |
| Haiku 4.5 | $0.00009 | $0.00322 |
Grade A, and why
world-model-diagnostic scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 5d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 305 lines — stays where its author put it; the contents beside it link to each section on GitHub.
World Model Diagnostic
Source: Jonathan Edwards (OB1 community), adapted for ai-agents.
Purpose
Your job is not to hand back a polished readiness score. Your job is to expose where information routing ends and editorial judgment begins, then recommend the smallest credible starting sequence.
This diagnostic answers five questions:
- Where does reality leave the clearest fingerprint in this business?
- Which world-model paradigm fits the company right now?
- Does the company have an explicit boundary layer?
- Where is it most exposed to simulated judgment?
- What should it build first, second, and third?
Triggers
| Trigger phrase | Operation |
|---|---|
run the world model diagnostic |
Start the 20-minute structured audit |
audit our world model |
Same as above, conversational form |
which world model architecture fits us |
Map company to paradigm |
audit where we automate judgment |
Boundary-layer audit only |
what should we build first for a world model |
Skip to recommended build sequence |
When to Use
Use this skill when:
- A team is choosing knowledge infrastructure (vector DB, ontology, telemetry pipeline) and needs to validate paradigm fit before investing.
- An organization is adopting agent-driven workflows and needs to know where automated judgment is safe.
- Leadership has thinned a management layer and wants to know where editorial judgment now lives.
Use a different skill when:
- You need a per-PR code or design review. Use
analystorarchitect. - You need product strategy or roadmap prioritization. Use
roadmap. - The org has already picked a paradigm and needs implementation help. Skip to the relevant build skill.
Non-Negotiable Rules
- Do not give a numeric readiness score.
- Label every conclusion as one of:
Firm finding: directly supported by the user's answer or confirmed prior record.Inference: synthesis from available evidence.Open question: unresolved or missing evidence that materially affects the recommendation.
- Keep the boundary layer central. Database choice is downstream of boundary clarity.
- Start concrete, not abstract. Ask about recent information flows, recent decisions, recent misses.
- Force ranking when discussing signal. Ask the user to rank the top 3 to 5 sources by fidelity.
- Audit actual flows, not aspirational diagrams.
- Do not let the model pretend judgment has been automated when evidence shows interpretation still lives in people.
- Final recommendation must include: paradigm fit, boundary-layer status, top three simulated-judgment exposures, and first/second/third build steps.
- Facts and interpretations cannot be presented with the same voice.
- Stay lightweight. Batch questions so the session finishes in about 20 minutes.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 5d ago First seen · 305 lines · 94 tokens per session scan A 1d4618c507fa
world-model-diagnostic is a skill published in the GitHub repository rjmurillo/ai-agents (45 stars, last pushed today), licensed MIT. It adds 94 tokens to every session and 3,216 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.
Other skills, from other repositories
memory-lancedb
LanceDB-backed vector memory for high-volume embedding and retrieval workloads.
rag-architecture
Build retrieval-augmented generation systems that ground LLMs in your data.
rag-knowledge-base
A local document search system that lets you add PDFs, Word files, Excel files, Markdown, and text, then ask questions about their contents. It finds relevant passages using both meaning-based and keyword search, then prepares context for an AI model.
document-qa
A document question-and-answer tool for PDFs, Word files, text, Markdown, CSV, and Excel files. It finds relevant passages in one file or a whole folder and uses a language model to answer questions.
llm-evals
Architecture standards, evaluation metrics, cost budget controls, and security guardrails for LLM, RAG, and Agentic features.
extremerouter-embeddings
Generate vector embeddings via ExtremeRouter /v1/embeddings using OpenAI / Gemini / Mistral / Voyage / Nvidia / GitHub embedding models for RAG, semantic search, similarity. Use when the user wants embeddings, vectors, RAG, semantic search, or to embed text.