Use when the user names a coding or Claude Code task ("add OAuth", "set up hooks", "review this PR", "pick a model"), asks which prompt to use, says to "use the right prompt", or when you enter this repo and need to orient without reading every file.
中文简介正在准备中。当前显示原文说明。
查看安装方法安装前可以查看交给 AI 代理的指令内容。
You route the current task to the right file(s) in this library, read them yourself, and then continue the existing work under those instructions — not from scratch, but picking up the conversation and any task already in progress.
${CLAUDE_SKILL_DIR} is this skill's own directory (<library-root>/.claude/skills/find-prompt/). The library root is ${CLAUDE_SKILL_DIR}/../../... All prompt files are under ${CLAUDE_SKILL_DIR}/../../../prompts/english/. Never hardcode an absolute path — always build it from ${CLAUDE_SKILL_DIR} so this works for anyone who clones the library, wherever they put it.
If ${CLAUDE_SKILL_DIR} is not substituted (it shows up literally), fall back: the skill sits next to SKILL.md; run ls up the tree from your working directory to find the folder containing prompts/english/agents/claude-agent-system-prompt.md.
Determine the task. Use $ARGUMENTS if given. Otherwise use the task already being worked on in this conversation (the user's last real request, the in-progress work, the files open). If genuinely unclear, ask one short question and stop.
Match the task to the routing table below. Pick the base prompt plus specialists by tier: 1 for a single-domain task (default), 2 only if the task spans two genuinely independent domains, or route to Multi-Agent Orchestration if the work needs isolation/review. See ${CLAUDE_SKILL_DIR}/../../../prompts/english/workflows/prompt-selector-guide.md for the tier definitions and the conflict-precedence rule if loaded prompts disagree.
Read the file(s) with the Read tool, using paths built from ${CLAUDE_SKILL_DIR}:
${CLAUDE_SKILL_DIR}/../../../prompts/english/agents/claude-agent-system-prompt.md${CLAUDE_SKILL_DIR}/../../../prompts/english/<path from the table>
If a path does not exist, tell the user to run /update-prompts and stop.Adopt them and continue. Treat the base prompt as your operating system prompt and the specialist as the domain protocol for this task. Then:
Confirm in one short block which prompts you loaded and what you're now doing, then proceed.
Paths are relative to prompts/english/.
| Task is about… | Load |
|---|---|
writing / debugging a skill, SKILL.md | agents/agent-skills-prompt.md |
| connecting a DB / browser / API / tool via MCP | agents/mcp-integration-prompt.md |
| a plugin, a marketplace, bundling/sharing config | agents/claude-code-plugins-prompt.md |
| a hook, running something on edit/commit/stop, blocking a command | agents/hooks-automation-prompt.md |
subagents, parallel agents, dynamic workflows, /batch, whole-codebase audit, writer/reviewer | agents/multi-agent-orchestration-prompt.md |
CLAUDE.md, settings.json, .claude/rules, permissions | agents/claude-code-workflow-prompt.md |
effort levels, ultrathink, plan mode, how much to think | agents/claude-code-modes-prompt.md |
| which model to use, Opus vs Sonnet vs Haiku vs Fable, pricing | workflows/model-selection-guide.md |
| what a current Claude Code build can do — plan mode, rewind, headless, surfaces | workflows/claude-code-native-features-guide.md |
building an agent in Python/TS, the Agent SDK, query() | workflows/agent-sdk-guide.md |
| AGENTS.md, one config across Cursor/Codex/Aider, the primary source for a claim | workflows/reference-resources.md |
| evaluating or installing a skill, plugin, MCP server, or agent toolkit | workflows/ecosystem-discovery-guide.md |
| Task is about… | Specialist |
|---|---|
| web frontend — React/Vue/Angular/Svelte, a page, a component | project-types/web-development-prompt.md |
| a REST/GraphQL/gRPC API, an endpoint, a contract | project-types/api-development-prompt.md (+ agents/api-design-graphql-prompt.md for schema-first design) |
| an end-to-end app, frontend + backend together | agents/fullstack-development-prompt.md |
| iOS / Android / React Native / Flutter / KMP | project-types/mobile-development-prompt.md |
| a desktop app — Tauri / Electron / MAUI / Qt | project-types/desktop-development-prompt.md |
| a data pipeline — ETL, streaming, Kafka, Airflow, dbt | agents/data-engineering-prompt.md |
| an ML model, training, RAG, an LLM feature | project-types/data-science-ml-prompt.md (+ agents/ai-llm-integration-prompt.md for app-side integration) |
| Kubernetes, CI/CD, Terraform/OpenTofu, Docker, Cloudflare | project-types/devops-cicd-prompt.md (+ agents/cloud-infrastructure-prompt.md for multi-region / IaC depth) |
| a database schema, SQL, indexing, a slow query | agents/database-optimization-prompt.md (+ project-types/database-sql-prompt.md for data modeling) |
| a game — Unity / Unreal / Godot / Bevy, netcode | project-types/game-development-prompt.md |
| firmware, embedded, IoT, an MCU, an RTOS | project-types/embedded-iot-prompt.md |
| a smart contract, Solidity, web3, an L2 | project-types/blockchain-web3-prompt.md |
| anything else / language-agnostic | project-types/general-software-development-prompt.md |
| Task is about… | Specialist |
|---|---|
| reviewing a PR or change set | agents/code-review-prompt.md. Also load agents/ui-design-systems-prompt.md if the change touches UI, or agents/security-audit-prompt.md if it touches auth/data. If git is available and you can see the diff, /code-review (the bundled skill) is an option too. |
| security — a vulnerability, threat model, auth, secrets | agents/security-audit-prompt.md |
| a production incident, "it's broken", root cause, a bug | agents/debugging-troubleshooting-prompt.md |
| reducing complexity or technical debt, cleaning up | agents/refactoring-prompt.md |
| adding tests, a test strategy, coverage, a flaky test | agents/testing-strategies-prompt.md |
| slowness — latency, throughput, cost, memory, bundle size | agents/performance-optimization-prompt.md |
| migrating a framework/runtime/DB, a major version bump | agents/migration-upgrade-prompt.md |
| logs, metrics, traces, alerting, observability | agents/monitoring-observability-prompt.md |
| fault tolerance — retries, circuit breakers, resilience | agents/error-handling-resilience-prompt.md |
| accessibility — WCAG, screen readers, a11y | agents/accessibility-audit-prompt.md |
| Task is about… | Specialist |
|---|---|
| system design, an architecture pattern, trade-offs | agents/architecture-patterns-prompt.md |
| choosing a tool or library — "what should I use for X" | agents/technology-stack-prompt.md |
| GDPR / HIPAA / SOC 2 / PCI, regulated scope | agents/compliance-governance-prompt.md |
| branching, commits, the release process | agents/git-version-control-prompt.md |
| DX, linting, onboarding, the dev environment | agents/developer-experience-tooling-prompt.md |
| a monorepo, multi-package coordination, Turborepo/Nx | agents/monorepo-complex-projects-prompt.md |
| writing docs, a README, API reference | agents/documentation-prompt.md |
| a design system, tokens, a component library, theming | agents/ui-design-systems-prompt.md |
| consequential research, source verification, product/service comparison, recommendation beyond coding | agents/evidence-driven-research-prompt.md |
agents/claude-agent-system-prompt.md, except for a task purely about running Claude Code (then the single Claude Code prompt is enough).${CLAUDE_SKILL_DIR}/../../../prompts/english/INDEX.md and route from its task table.Loaded: agents/claude-agent-system-prompt.md + <specialist path> [+ <optional second>]
Task (continuing): <one line, from the conversation>
Success criteria: <measurable>
Constraints carried over: <list, or "none">
Then proceed with the task under the loaded protocol — from where it stands, not from zero.
还没有评价。欢迎分享你的使用感受。
无需登录即可发表。欢迎分享使用感受、提问，或说明在哪些环境下无法运行。
还没有评论。来写第一条吧。
Use when evaluating, comparing, installing, or recommending a Claude Code plugin, Agent Skill, MCP server, agent toolkit, or GitHub-hosted AI extension, especially when popularity is weak evidence or the candidate requests shell, network, filesystem, or credential access.
中文简介正在准备中。当前显示原文说明。
Use when cutting a release, writing a CHANGELOG entry, or summarizing what changed since a tag — when the repo's commits follow (or mostly follow) Conventional Commits and a mechanically-accurate section breakdown beats a model's reading of the diff.
中文简介正在准备中。当前显示原文说明。
Use before committing, opening a PR, or reporting a task done — when you want ground-truth confirmation there are no conflict markers, live-looking credentials, debug leftovers, untracked TODOs, or oversized files, without spending a model turn re-reading every file yourself.
中文简介正在准备中。当前显示原文说明。
Use before publishing documentation, merging a docs PR, or reporting a Markdown-heavy repo's navigation as complete — when you want ground-truth confirmation that internal links, heading anchors, and directory links resolve, without opening every file yourself.
中文简介正在准备中。当前显示原文说明。
Use when the user names a coding, Claude Code, research, documentation, or technology-selection task; asks which prompt to use; says to use the right prompt; or when an agent enters this repository and needs to orient without reading every file.
中文简介正在准备中。当前显示原文说明。
Use when a skill isn't triggering as expected, before adding a new skill (to check for a name collision), during periodic cleanup of an accumulated skills directory, or when asked to review skill quality — checks frontmatter validity, description quality against Skill Discovery Optimization conventions, body size, and cross-scope name collisions.
中文简介正在准备中。当前显示原文说明。