無料GitHub で公開

my-bulk-read

Delegate reading large or multiple files to the `bulk-read` command, which sends the files to a cheaper reader model and returns only its answer. Use when a question needs the content of a file over about 350 lines or of several files at once, when a PreToolUse guard denied an unbounded read, or when the user says "bulk read", "一括読み込み", or "まとめて読んで". Do NOT use for files you are about to edit, for subtle correctness reasoning such as concurrency bugs, or for short files that fit a bounded read.

Ynakatsuka/dotfiles

インストール方法を見る

含まれるファイル（1）

SKILL.md3.4 KB

SKILL.md（原文）

インストールする前に、エージェントに与えられる指示の中身を確認できます。

Bulk Read Delegation

When to delegate

A question about what a file or module does, where something is defined, or how a flow works, and the answer does not need exact line numbers.
Several related files must be understood together.
The guard denied an unbounded read of a large file.

Keep reading directly with a bounded range (offset and limit, sed -n 'A,Bp') or a search (rg -n) when you need exact lines for an edit or a search can locate the section.

Command

bulk-read --question "<one precise question>" <file>...

Ask one specific question per call and name the identifiers or behavior you need. Vague questions return vague summaries.
Pass only the files the question needs. Each file is sent verbatim.
The answer arrives on stdout; startup information goes to stderr. Runtime depends on input size and reasoning. Silence is not evidence of completion or failure. A non-zero exit means the delegation failed. Report the error instead of retrying with the same input.

Collecting completion

This command is a shell process, not a spawn_agent task. Do not use wait_agent, clock.sleep, or shell sleep to wait for its result, and do not rely on a completion notification.
In Codex, when exec_command returns a session_id, collect output with write_stdin using that ID and empty chars. Use yield_time_ms: 1000 while waiting for this result; repeat while a session_id remains, until an exit_code and the final output arrive. The tool may enforce a longer minimum wait, but returns early when the process finishes.
functions.exec saying Script completed only ends the JavaScript wrapper. An inner session_id still needs collection. If the wrapper instead yields a running cell_id, resume it with functions.wait and yield_time_ms: 1000, then inspect the inner result.
Do independent work while the reader runs, but check its result at the next tool boundary. When its answer is the only remaining dependency, keep collecting directly without an extra delay. Once the exit code and answer arrive, consume them and stop waiting.
In Claude Code, collect a background Bash task with TaskOutput using its returned task ID; a finished task's output should be read immediately rather than waiting for another notification.

Working with the answer

Treat the answer as a summary from another model. Confirm identifiers with a targeted search before relying on them.
Line numbers in the answer are not reliable. Locate edit points with a bounded read.
When the answer says the files lack the information, widen the file set or search instead of asking the same question again.

Limits

The reader cannot edit files or run commands.
Delegation does not replace careful reasoning about subtle bugs. Read those sections yourself.

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

ログインしてレビューを書く

同じリポジトリのスキル

概要と使いどころ

my-agent

無料日本語概要

Delegate tasks to an external CLI agent (OpenAI Codex, Google Gemini, or Claude Code) for second opinions or parallel execution. Use when the user explicitly mentions "codex", "gemini", or "claude" or asks to delegate work to one of them (e.g., "codexで実行", "codexに聞いて", "geminiで実行", "Geminiに聞いて", "claudeで実行", "Claudeに聞いて", "run with codex", "ask gemini", "ask claude"). Also use when the user mentions Claude's "fable" model alias for delegation or consultation (e.g., "fableで実行", "fableに相談", "Fableに聞いて", "ask fable"). Do NOT use for general coding tasks that don't mention codex, gemini, claude, or fable.