メニュー
Disable autocomplete and AI assistants for distraction-free coding practice. Toggle on before interviews, off after.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
| name | focus-mode |
| description | Disable autocomplete and AI assistants for distraction-free coding practice. Toggle on before interviews, off after. |
Disable autocomplete and AI coding assistants for authentic interview practice.
/focus-mode on
Save current settings to .interview/.focus-backup.json:
cp .vscode/settings.json .interview/.focus-backup.json 2>/dev/null || echo "{}" > .interview/.focus-backup.json
Disable autocomplete - Update .vscode/settings.json:
{
"editor.quickSuggestions": false,
"editor.suggestOnTriggerCharacters": false,
"editor.parameterHints.enabled": false,
"editor.wordBasedSuggestions": "off",
"editor.inlineSuggest.enabled": false,
"editor.acceptSuggestionOnCommitCharacter": false
}
List AI extensions to disable:
code --list-extensions | grep -iE "copilot|codeium|tabnine|codewhisperer|continue|supermaven|cursor"
Output instructions:
Focus Mode: ON
Autocomplete disabled in VS Code settings.
Please manually disable these AI extensions:
- [list detected extensions]
To disable: Cmd+Shift+P → "Extensions: Disable" → select extension
Run `/focus-mode off` when done to restore settings.
Create marker file: .interview/.focus-mode-active
/focus-mode off
Check if focus mode is active:
test -f .interview/.focus-mode-active
If not active: "Focus mode is not currently active."
Restore settings:
cp .interview/.focus-backup.json .vscode/settings.json
Remove marker: rm .interview/.focus-mode-active
Output:
Focus Mode: OFF
Settings restored from backup.
Remember to re-enable your AI extensions:
Cmd+Shift+P → "Extensions: Enable" → select extensions
/focus-mode status
Focus Mode: [ON/OFF]
[If ON]:
- Autocomplete: Disabled
- AI Extensions: Manual check required
[If OFF]:
- Settings: Normal
The /interview skill checks focus mode status:
/focus-mode offDrill LeetCode problems by pattern until you can recognize and solve them in under a minute. Medium/Hard only. 20 min per problem.
Multi-part, progressively-revealed build problems (KV store, rate limiter, ledger, scheduler, task queue...). Part 2/3/4 unlock only after the prior part passes, so each part builds on your earlier code. Backend/systems, language-agnostic. Generated fresh each session.
Build Python muscle memory for coding interviews. Short, targeted drills on Python data structures and idioms.
Full-day-style "working session" build problems, in the mold of modern AI-native / work-sample interviews (Stellic, Sierra, Cursor/Anysphere, Meta AI-enabled loops). One open-ended, over-scoped, realistic engineering task in a real domain, followed by a project walkthrough. Judgment over completion. Traps are latent in the data, never labeled.
Given a role (e.g. "Senior SWE at Google"), searches all 5 research sources for interview problems asked in the last 3 months, maps vague descriptions to the nearest LeetCode problem, and expands the search if coverage is thin.
Research company-specific interview problems from Glassdoor, LeetCode Forums, Hello Interview, LeetCode problem sets, and Blind. Every finding must have a source URL. Defaults to last 3 months.