菜单
Use automatically for Claude Code work involving files, tools, bash, edits, tests, long commands, failed tool calls, or unfamiliar repos. Improves tool choice, prevents loops, and keeps work high-quality without adding broad context.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
Use when setting up or repairing repo agent instructions for Claude Code, Codex, or both. Proposes durable AGENTS.md, CLAUDE.md, and project memory changes before writing unless setup was explicitly requested.
Use for backend routes, API design, validation, database interactions, auth boundaries, response shapes, typed clients, error models, and integration-safe changes.
Use for slow pages, large bundles, slow APIs, database performance, caching, rendering jank, long sessions, and production-readiness polish.
Use for high-end UI, SaaS dashboards, landing pages, product websites, mobile responsive UI, visual redesign, UI polish, empty/loading/error states, motion, typography, and design quality rescue.
Use for long-running projects, context compaction, resume/handoff work, multi-session builds, or when the user asks to preserve project state. Proposes small verified memory docs before creating them unless setup was explicitly requested.
Senior code review workflow. Use when reviewing changes, checking quality before commit, evaluating generated code, or finding bugs/security risks.
| name | Tool Coach |
| description | Use automatically for Claude Code work involving files, tools, bash, edits, tests, long commands, failed tool calls, or unfamiliar repos. Improves tool choice, prevents loops, and keeps work high-quality without adding broad context. |
| allowed-tools | Read, Bash |
Use this skill to choose tools deliberately and avoid wasteful or unsafe loops.
Bash with rg, rg --files, or find before broad reads.Read before Edit/MultiEdit.Edit for small precise changes; MultiEdit for several edits in one file; Write only for new files or full rewrites.Bash for targeted verification, not blind full-suite runs unless needed.tail, filters, or targeted rg/grep.When MCP servers are available, prefer the official/current server for the target service over recreating the integration through ad hoc web fetches. Use MCP for service-native context such as GitHub, databases, observability, or issue trackers when it is authenticated and relevant. Do not use MCP for local filesystem work that the current host tools already handle better. If MCP auth is missing, state that clearly and continue with safe local evidence.
This LTS uses Bash for search because some Claude Code sessions do not expose direct Grep/Glob tools. For text search, call Bash with rg or command-line grep. For file discovery, call Bash with rg --files or find. Never emit tool_use.name = "Grep", tool_use.name = "Glob", tool_use.name = "grep", or tool_use.name = "glob".
Prefer:
mkdir -p /tmp/claude-lts-jobs
command > /tmp/claude-lts-jobs/<name>.log 2>&1 &
echo $!
Then check:
ps -p <pid> -o pid,etime,command
tail -n 80 /tmp/claude-lts-jobs/<name>.log
Use this only when background execution is actually useful. Otherwise run normally and wait.
Do not reduce correctness, tests, or security to save tokens. Reduce waste by narrowing context and running focused checks.