L’interface est localisée ; les présentations et documents des Skills sont traduits automatiquement par IA dans votre langue. Le code, les commandes, chemins et termes techniques restent en version originale, et le document de l’auteur fait foi.
Explorer les Skills
Données et recherche · SKILL.md

stan_model_contract

stan_model_contract

Extract a structured contract from simple Stan programs so score-model workflows can identify data, constrained parameters, transforms, model terms, and generated quantities.

Independent GitHub snapshot · Automated classification · Original documentation governs

Documentation de la Skill

GitHub

name: stan_model_contract description: Extract a structured contract from simple Stan programs so score-model workflows can identify data, constrained parameters, transforms, model terms, and generated quantities.

Stan Model Contract

Use this skill when a recovery or algorithm needs a lightweight, language-neutral contract for a small Stan model. Do not use it as a general Stan compiler or full parser.

Inputs

  • Stan source text or a path to a .stan file.
  • Optional observed data dictionary for downstream validation.

Outputs

  • JSON with data, parameters, transformed_parameters, model_terms, and generated_quantities.
  • Diagnostics for unsupported or missing sections.

Workflow

  1. Read the Stan source.
  2. Call scripts/stan_contract.py --stan <path> --output <json>.
  3. Inspect the output contract before passing it to transform or score-evaluation modules.
  4. Treat unsupported syntax as a blocker for this reduced parser, not as evidence about real BridgeStan.

Validation

Run python tests/test_stan_contract.py from this skill directory.

Limitations

The parser intentionally supports the Bernoulli-style recovery subset: scalar bounded real parameters, integer data arrays, simple transformed-parameter assignments, distribution statements, and generated-quantity assignments.

Skills enregistrées