mcpbeat Sign in

Curating Library Kg Agent Skill

Guide users through curating the library knowledge index — reviewing block categories, marking common/important blocks, and improving block descriptions via agent inference for better agent block selection.

2k tokens
context cost
the whole folder, loaded on every use
2
files
instructions only
0
copies elsewhere
how many repositories repackaged it
900
stars on the repo
on the repository, not the skill itself

Install

one command, takes just this skill from the repository
npx skills add https://github.com/matlab/simulink-agentic-toolkit --skill curating-library-kg

The instruction itself

14 sections, as written by the author

Curating the Library Knowledge Index

Curate .satk/library-kg/ for better agent block selection. Infer categories and descriptions from block metadata in .satk/library-cache/*.json, save curation to .satk/library-curation.json, then regenerate the KG.

When to Use

  • User wants to mark blocks as commonly used or important
  • User wants to correct auto-assigned categories
  • User wants to improve block descriptions for better selection
  • User asks "how do I make the agent prefer certain blocks?"
  • Called from building-simulink-models Gate 3.

When NOT to Use

  • Blocking, deprecating, or protecting blocks → configuring-block-policy
  • Actively building a model → building-simulink-models
  • Declaring which libraries exist → Library Setup gate in building-simulink-models

Prerequisites

  • satk-libraries.json or .satk/reuse-libraries.json must exist with libraries declared

Curation Rules

  • Custom descriptions are mandatory for ALL metadataQuality: "low" and "medium" blocks across ALL declared libraries. These blocks have no useful information. Write a short intent-focused description for every such block to help with agent block selection. Do not skip blocks because a file is large or because they seem unrelated to the current task.
  • Custom descriptions for "high" quality blocks are encouraged but optional.
  • Write custom descriptions for as many blocks as possible. After covering low/medium, covering high blocks too improves the KG.
  • Process in batches by category. Group blocks, infer descriptions together, present to user in digestible chunks.
  • Never write template or mechanical descriptions. Descriptions like "Mathematical operation: gain" or "Signal routing utility: mux" are useless — they just restate the category and name. Every description MUST be thoughtful, intent-focused description explaining *when* to use the block and *what modeling problem* it solves (e.g., "Scale a signal by a constant factor — use for unit conversion, controller gains, or applying physical constants" instead of "Mathematical operation: gain").

Modes

  • Automatic (from Gate 3 "Automatic" option) — infer everything, commit without pausing per step, present final result for review.
  • Guided (from Gate 3 "Guided setup" or direct user invocation) — propose at each step, wait for user confirmation before proceeding.

Determine mode before starting. Do not proceed without mode selection.

Workflow

If .satk/library-kg/index.md already exists, start at step 1 (review existing state). If not, start at step 2.

  • Review (only if KG exists) — Read index.md and common.md, summarize current state to user (libraries, block count, categories, common blocks).
  • Mode selection — Determine Automatic or Guided. Do not proceed without completing this step.
  • Descriptions — Ensure cache exists (see API), then read ALL .satk/library-cache/*.json files in full. If a file exceeds read limits, read it in chunks until every block has been processed. Apply Curation Rules above — write custom descriptions for blocks, prioritizing low/medium quality. Do NOT proceed to Step 4 until descriptions exist for every low/medium quality block across ALL cache files. Save to customDescriptions.
  • Common blocks — Propose which blocks should be marked as commonly used. Consider blocks covering diverse categories and frequent modeling workflows. Save to commonBlocks.
  • Categories — Finalize category definitions (names, descriptions, 3-5 keywords each). Allow user to correct individual block assignments via categoryAssignments.
  • Completeness check — Before saving, count total low/medium quality blocks across ALL cache files and count how many have custom descriptions. Report coverage to the user (e.g., "275/275 low/medium blocks described, 120/285 high blocks described"). Do NOT proceed if coverage of low/medium blocks is below 100%.
  • Save and generate — Save all curation data via library.LibraryCuration.save(projectRoot, curation), then run library.kg.Populate.run(projectRoot). Present the output summary to the user.

API

Ensuring cache exists

libConfig = library.LibraryConfig.load(projectRoot);
library.LibraryCatalog.getOrCreate(libConfig, projectRoot);

Reading block metadata

Read .satk/library-cache/*.json directly. Each file:

{
  "libraryName": "MotorLib",
  "description": "Motor control library",
  "blocks": [
    {
      "name": "SpeedController",
      "maskType": "SpeedCtrl",
      "blockType": "SubSystem",
      "maskDescription": "Closed-loop speed regulation with anti-windup",
      "description": "",
      "pathCategory": "Controllers",
      "metadataQuality": "high",
      "referenceBlock": "MotorLib/Controllers/SpeedController"
    }
  ]
}

Saving curation data

projectRoot = prefdir();
curation = library.LibraryCuration.load(projectRoot);
curation.commonBlocks = {'Speed Controller', 'Torque Estimator'};
curation.categories = struct('name', 'motors', 'description', 'Electric motors', 'keywords', {{'motor', 'drive'}});

% Use containers.Map — supports any block name as key
curation.customDescriptions = containers.Map('KeyType', 'char', 'ValueType', 'char');
curation.customDescriptions('Unit Delay') = 'Delay signal by one sample period';
curation.customDescriptions('1-D Lookup Table') = 'Interpolate output from breakpoint-value pairs';

curation.categoryAssignments = containers.Map('KeyType', 'char', 'ValueType', 'char');
curation.categoryAssignments('DC Current Controller') = 'motor-control';

library.LibraryCuration.save(projectRoot, curation);
library.kg.Populate.run(projectRoot);

Important: Use containers.Map (not struct) for customDescriptions and categoryAssignments. Struct field names cannot contain spaces or hyphens, which most Simulink block names have.

JSON format on disk

{
  "customDescriptions": [
    {"block": "Unit Delay", "value": "Delay signal by one sample period"},
    {"block": "1-D Lookup Table", "value": "Interpolate output from breakpoint-value pairs"}
  ],
  "categoryAssignments": [
    {"block": "DC Current Controller", "value": "motor-control"}
  ]
}

Curation Fields

| Field | Type | Effect |

|-------|------|--------|

| commonBlocks | cell array of strings | Always shown in common.md regardless of quality score |

| categories | struct array with .name, .description, .keywords | Defines categories with keyword matching for assignment |

| categoryAssignments | containers.Map (blockName → categoryName) | Per-block category assignment |

| customDescriptions | containers.Map (blockName → description) | Per-block custom description (intent) |

Guardrails

  • Never call find_system, get_param on library .slx files. Read .satk/library-cache/*.json for metadata, use library.kg.Populate.run() to generate the KG, and library.kg.Query.search() for lookups.
  • Never modify .satk/library-cache/*.json or .satk/library-kg/*.md directly — they are auto-generated.
  • Persist curation via library.LibraryCuration.save() only.
  • Confirm changes with the user before saving.
  • In user-facing output, do not use the term 'override'. Use 'custom description' or 'category assignment' instead.

anges with the user before saving

  • For progress tracking during curation, do not mention the term 'override' in user-facing output. Use 'custom description' or 'category assignment' instead.

----

Copyright 2026 The MathWorks, Inc.

----

Other skills for the same job

different authors, same section of the catalogue
At Dispatch V2
by pytorch
vendor ×3

Convert PyTorch AT_DISPATCH macros to AT_DISPATCH_V2 format in ATen C++ code. Use when porting AT_DISPATCH_ALL_TYPES_AND*, AT_DISPATCH_FLOATING_TYPES*, or other dispatch macros to the new v2 API. For ATen kernel files, CUDA kernels, and native operator implementations.

2k tokens
Docstring
by pytorch
vendor ×3

Write docstrings for PyTorch functions and methods following PyTorch conventions. Use when writing or updating docstrings in PyTorch code.

3k tokens
Statsmodels
by christophacham
×3

Statistical models library for Python. Use when you need specific model classes (OLS, GLM, mixed models, ARIMA) with detailed diagnostics, residuals, and inference. Best for econometrics, time series, rigorous inference with coefficient tables. For guided statistical test selection with APA reporting use statistical-analysis.

27k tokens
AI SDK
by vercel-labs
vendor ×2

Answer questions about the AI SDK and help build AI-powered features. Use when developers: (1) Ask about AI SDK functions like generateText, streamText, ToolLoopAgent, embed, or tools, (2) Want to build AI agents, chatbots, RAG systems, or text generation features, (3) Have questions about AI providers (OpenAI, Anthropic, Google, etc.), streaming, tool calling, structured output, or embeddings, (4) Use React hooks like useChat or useCompletion. Triggers on: "AI SDK", "Vercel AI SDK", "generateText", "streamText", "add AI to my app", "build an agent", "tool calling", "structured output", "useChat".

6k tokens
Create Llms
by github
vendor ×1

Create an llms.txt file from scratch based on repository structure following the llms.txt specification at https://llmstxt.org/

2k tokens
Esm
by K-Dense-AI
×1

Use when working directly with the `esm` Python SDK, ESM3 or ESMC model IDs, Forge/Biohub inference clients, or ESMFold2 folding workflows.

21k tokens
Modal
by K-Dense-AI
×1

Modal is a serverless cloud platform for running Python on demand, including on-demand GPUs. Use when deploying or serving AI/ML models, running GPU-accelerated workloads (training, fine-tuning, inference), serving web endpoints, scheduling batch jobs, or scaling Python code to cloud containers with the Modal SDK.

19k tokens
Pytdc
by K-Dense-AI
×1

Use Therapeutics Data Commons through the PyTDC Python package for registry discovery, approved dataset access, task-aware splits, evaluator metrics, benchmark groups, and bounded molecular-oracle workflows.

27k tokens scripts

How to use it

Copy the folder

Take matlab/curating-library-kg from the repository into ~/.claude/skills for personal use, or into .claude/skills inside a project.

Check the name does not clash

The agent identifies a skill by the name field in its header. Two skills with the same name cannot sit side by side — one of them will be ignored.