SuperAPI AI Directory

zio-knowledge

Coding & development

View repository

Updated 10-16 01:27

About this skill

ZIO Knowledge

Core Principles

  1. Accuracy over memory — Do not rely on training data for ZIO specifics. Fetch the relevant documentation page before answering.
  2. MCP tools first, sitemap fallback — If the zio-docs MCP server is connected, use mcp tools instead of WebFetch. Otherwise start at https://zio.dev/llms.txt to discover the current documentation structure and pick the right page for your specific question. Not connected? See https://zio.dev/start.md to add it.
  3. Source everything — Include the documentation URL (or doc path, when using the MCP tools) in your response so the user can verify and learn more.
  4. Right resource first — Navigate to the specific reference page rather than answering from the generic overview.
  5. Prefer Markdown pages — When falling back to WebFetch, always fetch the .md version of a documentation URL first. Fall back to the regular HTML website page only if the .md fetch fails or returns an error.

Question Routing

Any ZIO question?

→ Check whether the zio-docs MCP server (mcp.zio.dev/mcp) is connected:

  • MCP available: Call search_docs with the question first. Each result has a path (verified against the local index) or a source_url (external, fetch directly). Fetch the full page with get_doc_page(path). Use get_doc_index to browse all indexed pages when you need to discover paths rather than search by topic.

  • MCP unavailable: Fall back to the LLM sitemap:

    The sitemap follows the llmstxt.org standard and lists every documentation page with its URL and a one-line description. Read it to identify the most relevant page(s) for your question, then fetch those pages for API details, types, and method signatures. The URLs end with .md, which means they are Markdown files, and you can read their raw content.


Response Workflow

  1. Identify the topic — effects? fibers? concurrency? streams? testing? dependency injection? error handling? one of the ecosystem libraries?
  2. Route, fetch, and answer — do not answer from memory. Follow Question Routing above (MCP tools first, sitemap fallback, applying Core Principle 5 for the .md-first rule), then cite the source per Core Principle 3.
  3. If uncertain — direct the user to the official docs: "For the most current information, see https://zio.dev"

Quick Reference

MCP Server (prefer this when connected): https://mcp.zio.dev/mcp

LLM Sitemap (fallback, if MCP is not connected): https://zio.dev/llms.txt

Full Text File Documentation (single file): https://zio.dev/llms-full.txt — the concatenated content of every documentation page. Use it to index the full docs locally in one request when a session needs multiple questions answered across sections, instead of fetching pages individually.

Official Documentation: https://zio.dev

GitHub Repository: https://github.com/zio/zio

Maven Central: https://central.sonatype.com/artifact/dev.zio/zio_3

Examples Directory (GitHub): https://github.com/zio/zio/tree/series/2.x/examples


Common Failures

search_docs returns source_url not path — fetch source_url directly, don't call get_doc_page.

get_doc_page fails for a returned path — retry once verbatim, else fall back to sitemap.

MCP unavailable — fall back to sitemap workflow above.

Sitemap fetch fails — Retry once; if it persists, fall back to https://github.com/zio/zio (the source is the ground truth) and the README.

WebFetch returns 404 for a zio.dev/... page — it was renamed or removed. Re-fetch the sitemap (https://zio.dev/llms.txt), search for the topic, and navigate to the new URL.

Topic isn't in the sitemap — either it's too narrow for its own page or covered under a broader one. Search the examples directory and the source under https://github.com/zio/zio/tree/main/zio/src/main/scala/zio.

Multiple pages cover the same topic — prefer the page under zio.dev over external links; cite both if they materially differ.

Doc page contradicts your training data — Training cutoff is older than the docs site. Trust the docs site, not training data. Cite the page in the answer so the user can verify.