Skip to main content Skills Marketplace Descubre y explora habilidades de IA creadas por la comunidad.
Instalar con Codex o Claude Copia este prompt, pégalo en Codex, Claude u otro asistente, y deja que revise la página de la skill y la instale por ti.
Mostrar detalles del prompt Un comando directo omite el prompt de revisión. Revisa el origen antes de ejecutarlo.
npx skills add https://github.com/SimHacker/moollm --skill schema-factoryEl comando permanece en una sola línea. Desplázate horizontalmente para revisarlo antes de copiarlo.
¿Prefieres una copia local? Descarga los archivos que SkillsMP tiene disponibles ahora.
Descargar Zip Descargando... Explorador de archivos
10 archivos skill-snitch-report.md 8.1 KB Ocupaciones relacionadas SOC
Basado en la clasificación ocupacional SOC
name schema-factory description Build, lint, ingest, compose Drescher-style schemas license MIT tier 2 allowed-tools ["read_file","write_file","shell"] related ["schema-mechanism","experiment","debugging","planning"] tags ["moollm","schemas","drescher","deterministic","reasoning"] credits ["Gary Drescher — Schema Mechanism (1991)","Henry Minsky — Blocksworld examples"]
SCHEMA-FACTORY
"Deterministic checks first, LLM second."
Build, lint, ingest, compose, and generate context for Drescher-style schemas.
Why This Exists Gary Drescher's schema mechanism is strongest when it has:
Prescriptive schema-schema — what valid schemas must look likeDeterministic layer — evaluate and refine before asking the LLMContext generator — emit only needed patterns and evidence The goal is hybrid orchestration : Python does deterministic work, Cursor/LLM handles synthesis, MOOLLM stays explicit about what came from where.
Key Files File Purpose SCHEMA-SCHEMA.ymlDrives linting and ingestion schema_tool.pyCLI for all operations examples/schema-example.ymlCompact schema set examples/henry-minsky-blocksworld.ymlClassic microworld data
Quick Use
python3 schema_tool.py lint examples/schema-example.yml
python3 schema_tool.py compose --schemas examples/schema-example.yml --goal postgres-running
python3 schema_tool.py context --schemas examples/schema-example.yml --goal pyvision-running
Methods
LINT Validate schema against schema-schema.
Input: One or more schema files (YAML)Output: Pass/fail + diagnosticsEmits: schema_lintChecks: Required fields, type validation, reliability range, non-empty context/result
python3 schema_tool.py lint my-schemas.yml
INGEST Update schemas from experience logs or observed transitions.
Input: Experience logsOutput: Updated schema set + evidence countsEmits: schema_ingestDeterministic: No LLM calls; only schema updates
python3 schema_tool.py ingest experience-log.yml --into my-schemas.yml
COMPOSE Build action chain toward goal.
Input: Schema set + goalOutput: Composed action chain + rationaleEmits: schema_compose
python3 schema_tool.py compose --schemas my-schemas.yml --goal target-state
CONTEXT Generate compact context bundle for LLM synthesis.
Input: Schema set + goal + optional focus itemsOutput: Compact context bundleEmits: context_generateIncludes: id, action, context, result, reliability, extended_context, extended_results
python3 schema_tool.py context --schemas my-schemas.yml --goal target-state --focus item1,item2
Schema Structure schema:
id: "unique-identifier"
action: "what-the-schema-does"
context:
- precondition-1
- precondition-2
result:
- postcondition-1
reliability: 0.85
extended_context: [... ]
extended_results: [... ]
evidence_count: 47
marginal_attribution: {... }
Principles
Deterministic checks first, LLM second — Python validates before synthesisEmit events for traceability — Know what happened wherePrefer small, explicit context bundles — Don't dump everythingSchema-schema can evolve — Via the same learning loop it governs
The Schema-Schema The SCHEMA-SCHEMA.yml defines what valid schemas must look like:
Required fields and their types
Reliability range constraints
Context/result non-empty rules
Extension field patterns
This is prescriptive — schemas that don't match get lint errors.
Integration with LLM The factory provides deterministic foundation for LLM reasoning:
schema_factory compose --goal postgres-running
"Here are the relevant schemas and a proposed action chain..."
"Based on these schemas, I recommend..."
schema_factory ingest --experience new-observations.yml
Dovetails With
"Validate structure. Compose plans. Generate context. Let the LLM shine where it should."
