Initialize data engineering projects with lightweight context stores and AGENTS.md skeletons.
npx skills add https://github.com/Datus-ai/Datus-agent --skill init-datus-ai
Please help me install this Agent Skill. Skill: init Source: https://github.com/Datus-ai/Datus-agent/tree/main/datus/resources/skills/init Command: npx skills add https://github.com/Datus-ai/Datus-agent --skill init-datus-ai
Setting up a new data project typically requires time-consuming, token-heavy generation of full context stores like vector-indexed semantic models and metrics, which is unnecessary for early-stage exploration and wastes resources on projects that may not mature.
Invoke the init skill with optional scope hints like specific tables, domains, or file globs to automatically generate your project's lightweight inventory and context stores in a single confirmation-free pass.
High-intent search queries and answers about installing and using this skill.
Lightweight project initialization skips expensive vector-indexed stores by generating an AGENTS.md inventory skeleton and file-based knowledge entries. This delivers a usable first-pass project map and basic context for downstream agents with minimal token cost.
Domain classification scans in-scope files and database metadata to automatically categorize assets into a multi-level business domain taxonomy. This creates a clear project map for downstream agents without requiring manual categorization or heavy context generation.
Running a lightweight project initialization writes an AGENTS.md inventory skeleton automatically. It infers project goals and in-scope datasources from repository context or user hints, scoping all work to the relevant assets in a single confirmation-free pass.
Yes, cheap context generation writes atomic business facts to markdown knowledge files and durable cross-session preferences to memory. This file-based approach provides minimum viable context for downstream agents without building expensive vector-indexed stores.
No, lightweight scope resolution defaults to the whole project when no hints are provided. You can optionally supply specific tables, domains, or file globs to narrow the scope, and it will automatically restrict work to those relevant assets.
You should avoid lightweight project initialization when your project requires full semantic models or vector-indexed metrics for advanced queries. This approach deliberately skips expensive context generation to keep token costs low for early-stage exploration.
julianromliGenerate root and per-package AGENTS.md templates for repository submodules.
GmwEnterpriseLocate feature files via pre-mapped route maps and route tasks.
Datus-aiBuild vector-indexed knowledge bases from project files, database metadata, and validated SQL corpora.
ramonclaudioCombine BM25 keyword and vector semantic search across indexed reference collections.
