modelica-skills

Plugin modelica-skills from marketplace faveric-modelica-skills in faveric/faveric-modelica-skills. Commands read from the repo's manifest — 4 days ago.

Skills (17)

• approve-lessons

  Review the learner's candidates for the library overlay one by one and approve, edit or reject each. Nothing enters the library without this step.

• calibrate

  Fit model parameters to data without hiding structural errors in the parameters and without validating on the fitted data.

• cite-source

  Register a published source for an equation, correlation or constant used in a model, with where it is in the source, its valid range, and a check that it was transcribed correctly. Use the moment one appears.

• credibility-record

  Create and update the record that travels with a model: what it is for, its limits, which checks passed, and what status it may claim. Detects when the record is out of date.

• diagnosing-simulation

  Step-by-step method for models that will not translate, start, or finish, or that run and give wrong results.

• implement

  Build a task: write the failing check first, then the smallest model change that passes it, citing sources, updating the record once and closing the task with a short conclusion.

• interview-failed-issue

  Question session for a closed issue whose Outcome is missed: diagnose why from the existing evidence, rank the hypotheses, and end in recover, accept or stop. Writes the failure note, never code.

• interview-model

  Question session that fixes what a model is for, its physics, ranges and sources before any Modelica is written. Updates the glossary and records decisions.

• learn-from-issue

  Use after an issue's conclusion is written: collect the issue's evidence and propose at most three lessons or library skills for the overlay. Proposes only; never installs.

• model-design

  House rules for writing Modelica components across any domain: small models, clear acausal interfaces, units everywhere, range guards, bidirectional causality, no magic numbers. Use when writing or reviewing Modelica.

• review-model

  Review a finished issue in two independent passes (code standards, fidelity to the issue), check that the record does not claim more than the evidence, conclude the issue, run the learner and offer a PR.

• setup-modelica-skills

  One-time setup of a Modelica library repo for these skills: simulation engine (CLI/Docker or MCP server), where tasks, issues, references, and library overlays live.

• startsession

  Entry point for a working session: shows where each issue stands, asks what to do now, and suggests exactly one next command. Changes nothing.

• to-issue

  Turn the conversation into a model issue whose machine-readable header holds the facts (use, ranges, sources, tolerances, deliverables). Does not interview.

• to-tasks

  Split an issue into vertical slices, each adding one behaviour to the model as a complete mini-issue (problem, model, run, compare, answer) that one agent can open and close alone.

• verify-physics

  Run the model checks G1-G10, decide which status the evidence allows, and write results to the record. Use after any model change and before claiming a status.

• interviewing

  Interview the user about a plan or model, one question at a time with a recommended answer, until nothing is left open. Used by interview-model and calibrate.

MCP servers (2)

• modelica

  runs: uv run --quiet --with mcp>=1.2,<2 --with pyyaml python ${CLAUDE_PLUGIN_ROOT}/server/modelica_mcp.py

• dymola

  runs: uv run --quiet --with mcp>=1.2,<2 python ${CLAUDE_PLUGIN_ROOT}/server/dymola_mcp.py

Similar plugins