メニュー
Audit token costs of the .claude/ setup — skills, agents, hooks, and CLAUDE.md. Shows what's always loaded vs on-demand, flags bloat, and suggests cuts.
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
End-to-end verification - production build + banner generation tests (optional usernames, default nabondance)
Audit security-pinned dependency overrides in package.json. Checks whether each pinned version can be safely removed because the natural resolution now satisfies the minimum safe version. Triggered when the user asks to audit, review, or unlock pinned/locked dependencies, or check if overrides are still needed.
Test image generation API with optional username (default nabondance)
Validate production build
Start dev server in background
Stop background dev server
| name | context-budget |
| description | Audit token costs of the .claude/ setup — skills, agents, hooks, and CLAUDE.md. Shows what's always loaded vs on-demand, flags bloat, and suggests cuts. |
Run the audit script and report findings:
Execute: bash ../context-budget.sh
Then analyze the output:
Classify each item
/skill called)Flag if over budget
Report format
ALWAYS LOADED
CLAUDE.md ~XXX tokens
SessionStart.sh ~XX tokens (estimated output)
ON INVOCATION (skills)
/build ~XX tokens
/format ~XX tokens
...
ON AGENT SPAWN
graphql-explorer ~XXX tokens
TOTAL LOADED AT REST: ~XXX tokens
RECOMMENDATIONS: ...