Cantara

Use when refactoring, cleaning up, or enhancing the lib-electronic-components codebase. Provides guidance on architecture patterns, known issues, duplication hotspots, and recommended improvements.

Base skill for working with electronic components in this library. Use when adding new component types, manufacturer handlers, or working with MPN (Manufacturer Part Number) operations, BOM entries, or component classification.

Use this skill BEFORE creating a PR to ensure all documentation, skills, and learnings are updated. Critical for preserving institutional knowledge and preventing documentation drift.

Use when working with component similarity calculations - comparing MPNs, finding equivalent parts, implementing new similarity calculators, or understanding how component matching works.

Use when working with capacitor similarity calculations - comparing ceramic/electrolytic/film capacitor MPNs, understanding value/voltage/dielectric matching, or capacitor-specific similarity logic.

Use when working with connector similarity calculations - comparing connector MPNs, understanding pin count/pitch/family matching, or connector-specific similarity logic.

Use when working with diode similarity calculations - comparing signal/rectifier/zener/Schottky diode MPNs, understanding equivalent groups like 1N4148/1N914, or diode-specific similarity logic.

Use when working with LED similarity calculations - comparing LED MPNs, understanding color bins, brightness bins, families, or LED-specific similarity logic.

Use an existing KCP knowledge.yaml to find and load only the relevant knowledge for a task, instead of exploring a repository or docs site blind. Use when a project contains a knowledge.yaml (or a /.well-known/knowledge.yaml) and you need to answer a question or locate the right files efficiently.

Add a Knowledge Context Protocol (KCP) knowledge.yaml manifest to a codebase or documentation set so AI agents navigate it by intent instead of exploring blindly. Use when someone wants to make a repository "agent-ready", cut an agent's tool calls and context spend, adopt KCP, or upgrade an llms.txt to structured knowledge metadata.

Write or improve units in a KCP knowledge.yaml manifest — effective intents, triggers, relationships, audience targeting, negative-space (not_for), and temporal validity — and validate them. Use when editing a knowledge.yaml, when an agent keeps loading the wrong unit for a task, or when manifest queries return poor matches.

Safely ingest a third-party or untrusted KCP knowledge.yaml by running the trusted render pipeline (kcp render) before any of its content reaches the agent. Use when encountering a knowledge.yaml from a repository you do not control, an external/federated source, an unsigned manifest, or any manifest whose author you do not trust.

Manages local LLM models in Terteform. Use when asked to add, remove, activate, or deactivate a local model, download or start/stop a model server, register a new model in the registry, understand the models.conf / models.local.conf split, or how pi inside the container discovers local LLM servers.

Manages the Terteform Docker harness and local LLM servers. Use when asked to add domains to the whitelist, switch proxy modes, rebuild the image, manage terte containers, download/start/stop/health-check local LLMs, troubleshoot the Docker/OrbStack setup, or configure the mlx-lm or llama.cpp server.