claude-code-docs

name	claude-code-docs
description	Look up Claude Code documentation in the local Miyo-indexed folder labeled `claude-code`. Use when answering questions about Claude Code — features, configuration, hooks, permissions, settings, MCP, plugins, skills, sub-agents, Agent SDK, CLI flags, slash commands, env vars, cloud providers, or IDE integrations.
disable-model-invocation	true

Claude Code Documentation (local Miyo mirror)

A local Markdown mirror of the official Claude Code docs (~150 pages from code.claude.com, plus the release changelog), indexed in Miyo under the folder label claude-code. The docs are in English — phrase Miyo queries in English for best recall.

Reach the docs through Miyo's folder label claude-code

Every Miyo call below uses the folder label claude-code.

• Confirm the label with mcp__miyo__list_folders — you should see claude-code (… files, ready). If it was indexed under a different label, use that one everywhere below.
• Paths are Miyo-relative, of the form claude-code/<file>.md — exactly as mcp__miyo__search / mcp__miyo__list_files return them. Pass them straight to mcp__miyo__read_file; the native Read tool can't open them.

File naming

The mirror is a flat folder. Nested doc URLs flatten with __ as the separator:

• code.claude.com/docs/en/hooks → claude-code/hooks.md
• code.claude.com/docs/en/agent-sdk/skills → claude-code/agent-sdk__skills.md

So a whole topic family shares a filename prefix — every SDK page is agent-sdk__*.md.

How Miyo retrieval works (build better queries)

Miyo search is hybrid: for every query it runs two retrievers and fuses them with RRF (Reciprocal Rank Fusion), then applies a literal-match boost (an exact query substring in the title, then in the body, is pushed up):

• Dense (semantic) — embeds the query, matches by meaning. Good on paraphrase, weak when your wording is far from the docs' actual terms.
• BM25 (lexical) — keyword match. Good on exact terms/jargon, blind to synonyms.

A hit must surface in at least one retriever's prefetch to appear at all — RRF only ranks what was already pulled. So the usual failure is recall, not ranking: if neither semantic nor keyword catches it, raising limit or reranking won't help.

Query rules that follow from this:

• Use Claude Code's own vocabulary ("PreToolUse hook", "permission mode", "MCP server", "subagent", "settings.json", "output style") rather than a casual paraphrase — this feeds BM25 and aligns the dense vector simultaneously.
• Name the exact term you expect in the doc (a setting key, CLI flag, hook event, env var) — the literal boost rewards an exact substring.
• If the first query misses, reformulate with different/added domain terms or fire 2–3 variants; don't just bump limit.
• Do not prepend instruction-style prefixes (e.g. Instruct: …) — Miyo embeds the query text verbatim and symmetrically with documents, so that only adds noise.

Search the whole docset

Use Miyo semantic search scoped to the folder:

mcp__miyo__search(
  query: "<natural-language question>",
  folder_path: "claude-code",
  limit: 5            # 3–8 is usually enough; raise for broad topics
)

Search a specific topic

Add the path filter (case-insensitive substring on the result path) to narrow to one topic. The corpus is flat, so path matches a filename fragment — use the slug prefix, no slashes:

# Agent SDK pages only
mcp__miyo__search(query: "...", folder_path: "claude-code", path: "agent-sdk", limit: 5)

# Hooks
mcp__miyo__search(query: "...", folder_path: "claude-code", path: "hooks", limit: 5)

# MCP
mcp__miyo__search(query: "...", folder_path: "claude-code", path: "mcp", limit: 5)

⚠️ path must be a single fragment with no slash. Miyo stores the folder prefix with the OS-native separator (\ on Windows, / on macOS/Linux), so a filter containing a separator silently matches nothing on the wrong OS. Match the filename slug alone (e.g. path: "settings", path: "permissions").

Miyo returns ranked chunks with file paths and the matched snippet. For most questions the snippets answer directly — synthesize and cite source files inline by their Miyo-relative path, e.g. [hooks.md](claude-code/hooks.md).

When to read the full file

Fetch the raw file when the search chunk lacks the exact detail — an exact JSON schema, YAML frontmatter, precise CLI flag spelling, or a full env-var / settings / hook-event reference table.

Pass the Miyo-relative path that search/list_files returned straight to mcp__miyo__read_file — no path rewriting:

mcp__miyo__read_file(file_path: "claude-code/<file>.md")

Some Claude Code pages are large (the CLI, settings, and hooks references especially). mcp__miyo__read_file returns the whole file as one blob and can be truncated by the harness — if that happens, lean on the search chunks (raise limit, or add a path filter) to pull just the relevant sections instead.

Anti-patterns

• Do not call mcp__miyo__search without folder_path: "claude-code" — results will leak in from other indexed folders.
• Do not read raw docs first to "find" something — search first, read only when a chunk lacks a specific detail.
• Do not use WebFetch against code.claude.com — the local mirror is fresher and rate-limit-free.

Sub-commands the user may invoke

/claude-code-docs <question>

Search the whole docset (folder_path: "claude-code"). Synthesize an answer from the top 3–5 chunks and cite source files.

/claude-code-docs <topic> "<question>"

Search scoped to a path fragment — e.g. sdk → path: "agent-sdk", hooks → path: "hooks", mcp → path: "mcp" (single fragment, no slash).

/claude-code-docs explain "<concept>"

Broader search (limit: 10), group hits by file, and explain the concept covering: definition, configuration, related features, gotchas.

/claude-code-docs (no args)

List what's indexed (mcp__miyo__list_files(file_path: "claude-code/"), or mcp__miyo__list_folders for the count) and suggest the user phrase a question.