agentflow-read

name

agentflow-read

description

Read a single file and return ONLY the parts relevant to a query, instead of loading the whole file into the main session. Use when the user wants to check, find, or answer something about one specific file ("what does X do in foo.ts", "does this file handle Y", "show me the auth logic in handler.py") and the file is large enough that loading all of it would waste primary context. Dispatches a Haiku worker that reads the file in an isolated context and returns a focused, line-cited extract. For broad multi-file search use agentflow-search; for whole-file edits use the native Read tool directly.

agentflow-read

Return only the query-relevant parts of one file, keeping its full contents out of the primary session. A Haiku-pinned worker reads the file in a throwaway context and hands back a focused extract.

Honest note: native Read also works and is simpler for small files. Reach for this skill when the file is large, or when you want the cheaper isolated path.

When it fires

• "What does parseConfig do in src/config.ts?"
• "Does server.ts handle reconnects?"
• "Show me just the error-handling in this 2k-line log handler."

How to run it

Dispatch the agentflow-haiku-worker subagent (Agent/Task tool, model haiku) with the prompt below. Fill in {file_path}, {query}, and {mode} (relevant | summary | full; default relevant). Return the worker's output verbatim to the user.

You are a code reading assistant. Given a file and a query, return ONLY the parts
of the file that are relevant to the query.

STRICT RULES:
1. NO preamble. NO "Here is...". NO "Looking at the file...". Start with the result.
2. NO markdown fences around the output.
3. If the query is a yes/no or factual question that can be answered directly from
   the code, lead with a 1-2 sentence answer, then cite the supporting line(s)
   using the format "  42  | code line".
4. Otherwise, return ONLY the matching sections with minimal surrounding context
   (typically +/-3 lines) using the line-number format above.
5. Do NOT dump the entire file. Do NOT include unrelated code. If nothing matches,
   output exactly: NO_MATCH
6. Mode behavior:
   - "relevant" (default): only matching sections with line numbers
   - "summary": high-level overview of file structure and purpose, no source quotes
   - "full": complete file with brief inline annotations on relevant lines only

File: {file_path}
Mode: {mode}
Query: {query}

Read the file with the Read tool, then apply the rules above.

Other runtimes (Codex, and tools without Claude subagents)

The worker above is a Claude Code agent definition; elsewhere only the dispatch changes — the methodology prompt is identical:

• Codex CLI: spawn_agent with the filled prompt (requires multi_agent = true in ~/.codex/config.toml), wait_agent for the result, then close_agent.
• Other tools: use the runtime's isolated-subagent mechanism; if none exists, run the prompt inline and honor its output rules exactly.

No model pin exists outside Claude Code — pick the cheapest/fastest model if the runtime lets you choose. Since the AgentFlow worker agent is absent, prepend this line to the prompt: "You are a disposable worker in an isolated context. Follow the methodology exactly and return only the result — no preamble, no commentary."