worldforge CLAUDE.md

worldforge CLAUDE.md is an instructions file for coding agents from AbdelStark/worldforge. It costs 4,109 tokens per session, scanned A, original, MIT.

Repository instructions for WorldForge, a local framework for designing and evaluating workflows for robots and other physical AI systems.

In plain words
What is it for?
Use them when changing planning, scoring, provider, evaluation, or benchmarking code in WorldForge.
Why use it?
They keep the project focused, prevent unsupported dependencies, and define clear error handling and local-only boundaries.

Instructions file

Install

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.

agentmods
npx agentmods add instructions/abdelstark/worldforge/claude-md
Clone the repo
git clone --depth 1 https://github.com/AbdelStark/worldforge

Wrote 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.

agentmods badge for worldforge CLAUDE.md

README.md
[![agentmods](https://agentmods.dev/badge/instructions/abdelstark/worldforge/claude-md.svg)](https://agentmods.dev/instructions/abdelstark/worldforge/claude-md)
Your own site
<a href="https://agentmods.dev/instructions/abdelstark/worldforge/claude-md"><img src="https://agentmods.dev/badge/instructions/abdelstark/worldforge/claude-md.svg" alt="Measured on agentmods" height="20"></a>
Per session 4,109 This file is loaded in full into every session.
When invoked 4,109 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.04109 $0.04109
Opus 5 $0.02055 $0.02055
Sonnet 5 $0.00822 $0.00822
Haiku 4.5 $0.00411 $0.00411

Measured 4d ago against content hash 460126ddebb1, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

worldforge CLAUDE.md 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 4d 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.

CLAUDE.md · 233 lines

How it starts

The opening of the file, as written. The whole thing — 233 lines — stays where its author put it; the contents beside it link to each section on GitHub.

<priority_rules>

  1. Keep the scope narrow to the backbone loop: planning and scoring with an action-conditioned predictive world model in latent space, plus evaluation/benchmarking to select configurations. Do not reintroduce world generation, symbolic-scenario simulation, or media-generation features.
  2. Keep provider capabilities truthful: advertise only callable, tested, typed WorldForge surfaces.
  3. Make the library fault tolerant and developer-friendly while preserving the optional-runtime boundary: never add torch, LeWorldModel, LeRobot, GR00T, CUDA, checkpoints, datasets, or robot controllers to base dependencies or the repo.
  4. Fail loudly at boundaries: invalid public inputs raise WorldForgeError; malformed persisted/provider state raises WorldStateError; provider/runtime failures raise ProviderError.
  5. Preserve local-first scope: no hosted service, production database, credential store, robot safety layer, telemetry backend, or durable multi-writer persistence unless explicitly designed and approved.
  6. Public contribution artifacts must be human, maintainer-style, and tool-neutral. Do not mention agent/tool branding in branch names, commits, PR titles, PR bodies, changelog, docs, or README copy. </priority_rules>
Path Purpose Agent zone
src/worldforge/ Package source and public runtime Modify with tests; public API files are gated
src/worldforge/models.py Public models, validation, request policy, events, state contracts Gated for breaking contract changes
src/worldforge/framework.py WorldForge, provider orchestration, capability dispatch (predict/score/policy/embed), diagnostics, facade helpers Gated for public behavior changes
src/worldforge/providers/ Provider base classes, catalog, concrete adapters, optional runtimes Modify with provider skill and fixture tests
src/worldforge/evaluation/ Deterministic evaluation suites and renderers Modify with evaluation skill
src/worldforge/harness/ Robotics showcase flow/report package Keep Textual isolated to tui.py
src/worldforge/rerun.py Optional Rerun bridge Events, worlds, plans, benchmarks, robotics showcase visuals; keep out of base deps
src/worldforge/smoke/ Packaged optional-runtime smoke entry points Host-owned dependencies only
src/worldforge/testing/ Reusable adapter contract helpers Public testing API; modify carefully
tests/ Unit, contract, CLI, docs, fixture, regression tests Create/modify freely for changed behavior
tests/fixtures/providers/ Remote/provider parser fixtures Create/modify for provider failure modes
examples/ Runnable checkout examples and compatibility wrappers Keep deterministic unless explicitly live-smoke
docs/src/ User docs, architecture, playbooks, provider pages, API notes Update with public behavior
specs/ Per-feature spec triads (<feature>/{spec.md, plan.md, tasks.md}) following the GitHub Spec Kit pattern Author a new triad before implementing a multi-task feature; update the triad as scope changes
scripts/ Docs generator, provider scaffold, package check, smokes Gated for workflow/CI-impacting changes
.github/workflows/ CI, release, security pipelines Explicit approval before modifying
pyproject.toml, uv.lock Package metadata, deps, lockfile Explicit approval for dependency/version changes
README.md, CHANGELOG.md, CONTRIBUTING.md, AGENTS.md Public front face and canonical agent/multi-agent guide Update when public behavior or agent context changes; do not create lowercase agents.md duplicate
.env.example Tracked provider env template Modify only with provider env changes; never include secrets
.codex/skills/ Project-local agent skills Modify only when agent context is the task
.claude/skills, .agents/skills Symlinks to .codex/skills Keep as symlinks

Read the full file on GitHub · 233 lines

Changes

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.

  1. 4d ago First seen · 233 lines · 4,109 tokens per session scan A 460126ddebb1

Subscribe to this mod's changes

worldforge CLAUDE.md is an instructions file published in the GitHub repository AbdelStark/worldforge (108 stars, last pushed today), licensed MIT. It adds 4,109 tokens to every session, about $0.0205 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 2026-08-30.