本文へ移動
cccskills
無料GitHub で公開

file-editor

Use when opening edited files in the user's IDE. Reads settings from .agent-settings.yml to determine IDE and whether auto-open is enabled.

インストール方法を見る

含まれるファイル(1)

  • SKILL.md6.9 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。

<!-- cloud_safe: noop -->

file-editor

When to use

This skill applies automatically whenever:

  • The agent creates or edits a file using save-file or str-replace-editor
  • The agent wants to show the user a file they should review

Do NOT use when:

  • Only reading files (no edits)
  • Running commands or tests
  • The personal.open_edited_files setting is false

Procedure: Open files in IDE

  1. Read settings — Check personal.open_edited_files and personal.ide in .agent-settings.yml.
  2. Skip if disabled — If personal.open_edited_files is false, do nothing.
  3. Open files — Use the IDE CLI command to open each edited file.
  4. Verify — Confirm the command succeeded (exit code 0).

Settings file

Settings are stored in .agent-settings.yml (project root, git-ignored). See .augment/templates/agent-settings.md for the full settings reference.

Relevant settings for this skill:

KeyValuesDefaultDescription
personal.idecode, cursor, windsurf, phpstorm, webstorm, idea, goland, rubymine, pycharm, rider, subl, vim, nvim, emacs, zed(empty)CLI command to open files
personal.open_edited_filestrue, falsefalseWhether to auto-open edited files

Behavior

1. Read settings

At the start of a conversation (or on first file edit), read .agent-settings.yml:

cat .agent-settings.yml 2>/dev/null
  • If the file doesn't exist → do not open files. Suggest running scripts/install (then /onboard).
  • If personal.open_edited_files: false → do nothing.
  • If personal.open_edited_files: true and personal.ide is set → open files after edits.

2. Open files after edits

After editing one or more files, open them using the ide setting. Always jump to the first edited line. The syntax depends on the IDE:

JetBrains IDEs (personal.ide: phpstorm / webstorm / idea / pycharm / goland / rubymine / rider):

phpstorm --line {line} {file}        # PHP
webstorm --line {line} {file}        # JS / TS
idea     --line {line} {file}        # Java / Kotlin / general
pycharm  --line {line} {file}        # Python
goland   --line {line} {file}        # Go

VS Code-family (personal.ide: code / cursor / windsurf):

code    --goto {file}:{line}
cursor  --goto {file}:{line}
windsurf --goto {file}:{line}

Zed (personal.ide: zed):

zed {file}:{line}

Terminal editors (personal.ide: vim / nvim / emacs):

vim  +{line} {file}
nvim +{line} {file}
emacs +{line} {file}

For multiple files, run one command per file:

# JetBrains (PhpStorm / WebStorm / GoLand / PyCharm)
phpstorm --line 42 src/Services/UserService.php
webstorm --line 15 src/services/user-service.ts
goland   --line 27 internal/user/service.go
pycharm  --line 88 app/services/user_service.py

# VS Code-family
code     --goto src/services/user-service.ts:15
cursor   --goto app/services/user_service.py:88

# Open file without jumping to a line
code     src/services/user-service.ts
phpstorm src/Services/UserService.php

Rules

  • Batch opens — collect all files from a single logical edit operation and open them as one batch (one IDE command per file).
  • No duplicates — don't open the same file twice in one batch.
  • Silent on failure — if the IDE command fails, log it but don't interrupt the workflow.
  • Never open test output — don't open files just because they appeared in test results.
  • Only edited files — only open files the agent actively created or modified, not files that were only read.
  • Respect the setting — if personal.open_edited_files: false or .agent-settings.yml doesn't exist, never open files.

IDE CLI commands

IDECommandInstall
VS Codecode {file}Shell Command: Install 'code' in PATH
Cursorcursor {file}Settings → "Install 'cursor' shell command"
Windsurfwindsurf {file}Settings → "Install 'windsurf' shell command"
Zedzed {file}Settings → "Install CLI" (creates zed in PATH)
PhpStormphpstorm {file}JetBrains Toolbox CLI or Tools → Create command-line launcher in PhpStorm
WebStormwebstorm {file}JetBrains Toolbox CLI or Tools → Create command-line launcher in WebStorm
IntelliJ IDEAidea {file}JetBrains Toolbox CLI or Tools → Create command-line launcher in IDEA
PyCharmpycharm {file}JetBrains Toolbox CLI or Tools → Create command-line launcher in PyCharm
GoLandgoland {file}JetBrains Toolbox CLI or Tools → Create command-line launcher in GoLand
Vim / Neovimvim {file} / nvim {file}Bundled with most distros

Output format

  1. Files opened in the configured IDE (if enabled)
  2. No output if personal.open_edited_files is disabled

Auto-trigger keywords

  • open file
  • open in editor
  • show in IDE
  • file editor settings
  • configure IDE

Gotcha

  • Check .agent-settings.yml for personal.open_edited_files before opening anything — the user may have disabled it.
  • Don't open files during batch operations (e.g., fixing 20 type-checker errors across PHPStan / tsc / mypy / cargo check) — only open when specifically relevant.
  • JetBrains IDEs (PhpStorm, WebStorm, IDEA, PyCharm, GoLand, RubyMine, Rider) sometimes briefly lock a file on open — wait a moment before editing the same file.

Do NOT

  • Do NOT open files without checking .agent-settings.yml first.
  • Do NOT prompt the user about IDE settings during normal work — suggest /onboard (for first-run) or editing .agent-settings.yml directly.
  • Do NOT open files that were only read, not edited.
  • Do NOT open more than 10 files at once — summarize instead.

Cloud Behavior

On cloud surfaces (Claude.ai Web, Skills API) this skill is fully inert — there is no local IDE to open files in, no .agent-settings.yml to read, and no shell handler available. The cloud agent simply finishes its edits and reports back; file-opening is a local-agent concern.

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。

同じリポジトリのスキル

概要と使いどころ

Use when reviewing UI for accessibility — WCAG 2.2 AA, keyboard nav, focus, ARIA, contrast, screen-reader semantics — even on 'is this a11y-OK?' or 'mach das barrierefrei'.

日本語の概要は準備中です。原文の説明を表示しています。

event4u-app/agent-config112026年10月12日 更新

Use when defining or auditing the activation event — aha-moment selection, retention correlation, falsifiable definition. Triggers on 'what is our aha moment', 'redefine activation'.

日本語の概要は準備中です。原文の説明を表示しています。

event4u-app/agent-config112026年10月12日 更新

Use when capturing an architectural decision — file naming, next ADR number, Status / Context / Decision / Consequences, index regen; fires even without saying 'ADR'.

日本語の概要は準備中です。原文の説明を表示しています。

event4u-app/agent-config112026年10月12日 更新

Adversarial critique — devil's advocate, stress-test, honest teardown ('poke holes', 'be brutal', 'was hältst du davon'); explicit request only. Routine code or design review → code-review.

日本語の概要は準備中です。原文の説明を表示しています。

event4u-app/agent-config112026年10月12日 更新

Use when reading, creating, or updating agent documentation, module docs, roadmaps, or AGENTS.md. Understands the full .augment/, agents/, and copilot-instructions structure.

日本語の概要は準備中です。原文の説明を表示しています。

event4u-app/agent-config112026年10月12日 更新

Use for an adversarial red-team / blue-team / auditor review of an AI agent's CONFIG + behaviour (rules, skills, MCP, hooks, permissions) — attack-chain → defensive-gap list, not a code audit.

日本語の概要は準備中です。原文の説明を表示しています。

event4u-app/agent-config112026年10月12日 更新

event4u-app のスキルをすべて見る

このスキルの問題を報告する