agent-technology

Authors technology expert agent prompts with documentation source tables and companion sources files.

Updated Nov 7, 2023
One-click install
npx skills add https://github.com/lstellway/nixfiles --skill agent-technology-lstellway
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: agent-technology
Source: https://github.com/lstellway/nixfiles/tree/main/modules/home-manager/claude/skills/agent-technology
Command: npx skills add https://github.com/lstellway/nixfiles --skill agent-technology-lstellway

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Technology agents built on training data alone hallucinate stale API signatures, CLI flags, and configuration options. This Skill produces expert technology agents grounded in fetch-first discipline, pairing each agent with a verified documentation sources table. ## Core Features & Use Cases - Nine-step authoring workflow: Surveys the technology landscape, maps query types to authoritative URLs, separates stable embedded knowledge from volatile always-fetch knowledge, and defines per-task-type strategies. - Structural variants: Selects flat, partial, or full broad-surface structures based on whether the technology spans distinct sub-ecosystems (e.g., Docker vs. WordPress vs. Nix). - Dual output files: Writes the agent system prompt and a companion sources file recording version calibration, verified URLs, and design notes. - Use Case: Ask to create a Postgres expert agent and receive technology-postgres.md with a Context7-first documentation sources table, embedded core concepts, and a technology-postgres.sources.md provenance file. ## Quick Start Create a technology expert agent for Kubernetes with verified documentation sources and a companion sources file.

Frequently Asked Questions about agent-technology

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I create a technology expert agent for Claude?

Run the nine-step authoring workflow: survey the technology landscape, map query types to authoritative documentation URLs, embed stable core concepts, define per-task-type strategies, then write the agent file and a companion sources file to ~/.claude/agents/.

What is a documentation sources table in a technology agent?

It is a table mapping each query type (syntax reference, CLI flags, configuration options, changelogs) to the single best authoritative URL to fetch. It separates a technology agent from generic training-data answers, since APIs and options go stale quickly.

When should an agent use Context7 versus direct documentation URLs?

Context7 is the default top row for any indexed technology, providing version-pinned, LLM-targeted documentation snippets. Direct URLs are used as fallback, and Context7 becomes mandatory when doc sites are client-rendered and return title-only HTML to fetchers.

How do I choose between flat, partial, and broad-surface agent structures?

Use flat for single-ecosystem tools like Nix or Helm, partial when sources are distinct but tasks cross sub-domains like Docker, and full broad-surface when task flow itself branches by sub-domain, as with WordPress or Kubernetes.

What knowledge should be embedded versus always fetched in a technology agent?

Embed stable foundational concepts like data models and syntax primitives that hold across versions. Always fetch volatile knowledge like option lists, API signatures, package attributes, and CLI flags that change between releases.