跳到正文
cccskills
免费在 GitHub 公开

claude-code-custom-agents

Create and maintain Claude Code custom subagents (.claude/agents/*.md files). USE FOR: authoring project (.claude/agents/) or user (~/.claude/agents/) subagent files; frontmatter fields (name, description, tools, disallowedTools, model, permissionMode, maxTurns, skills, mcpServers, hooks, memory, background, effort, isolation, color, initialPrompt); writing descriptions that drive automatic delegation and 'use proactively' phrasing; invoking subagents via natural language, @-mention, session-wide --agent, or the Agent tool; restricting which subagent types a coordinator can spawn with Agent(name1, name2); nested subagents and depth limits; built-in subagents (Explore, Plan, general-purpose, claude); troubleshooting a subagent that won't load or isn't invoked. DO NOT USE FOR: GitHub Copilot CLI custom agents (use copilot-cli-custom-agents); agent teams/teammates (a different, session-spanning feature); MCP server authoring; SKILL.md authoring (use create-skill).

中文简介正在准备中。当前显示原文说明。

NetFabric/NetFabric.Numerics星标 36(GitHub 36 + cccskills 0)
集成工具GitHub

实测评分 未实测

查看安装方法

包含的文件(6)

  • SKILL.md6.1 KB
  • references/authoring-workflow.md3.4 KB
  • references/delegation-and-squads.md3.5 KB
  • references/frontmatter-reference.md5.0 KB
  • references/undocumented-and-gotchas.md3.3 KB
  • references/writing-style.md2.8 KB

SKILL.md(原文)

安装前可以查看交给 AI 代理的指令内容。

Claude Code Custom Agents

A subagent is a Markdown file (.md) with YAML frontmatter plus a system-prompt body. Only name and description are required. Claude Code matches your request (and its own reasoning) against every subagent's description to decide when to delegate — a vague description rarely gets used.

Anatomy

PartPurpose
Frontmattername + description (required) + optional tools, disallowedTools, model, permissionMode, maxTurns, skills, mcpServers, hooks, memory, background, effort, isolation, color, initialPrompt
BodySystem prompt: role, numbered workflow, output format. The subagent receives only this plus basic environment details (CLAUDE.md, git status) — not the full Claude Code system prompt
Location.claude/agents/ (project, walked to Git root) and ~/.claude/agents/ (user) are watched live, no restart needed for edits to an existing directory; managed settings and --agents CLI JSON outrank both; plugin agents/ is lowest priority

Minimal example

---
name: code-reviewer
description: Reviews code for quality, security, and best practices. Use proactively after code changes.
tools: Read, Grep, Glob
model: sonnet
---

You are a senior code reviewer. When invoked, run `git diff`, focus on
modified files, and provide feedback organized by priority (critical,
warnings, suggestions) with specific fix examples.

Invocation

MethodSyntax
Automatic delegationClaude matches your request against every subagent's description — include "use proactively" to encourage it
Natural languageName the subagent in your prompt; Claude decides whether to delegate
@-mention@agent-<name> — guarantees that specific subagent runs for one task
Session-wideclaude --agent <name> or "agent": "<name>" in .claude/settings.json — the whole session adopts that subagent's prompt, tools, and model
From another agentThe Agent tool (renamed from Task in v2.1.63) — nests up to 3 layers below the main conversation by default

Full frontmatter field table, tool-availability filters, and location priority → references/frontmatter-reference.md.

Built-in subagents

Explore (read-only search, model inherited/capped at Opus), Plan (plan-mode research), general-purpose (every subagent tool, for mixed research+edit tasks), claude (catch-all, also the default background-session agent), plus helper agents (statusline-setup, claude-code-guide) Claude invokes automatically. A user/project subagent named Explore overrides the built-in.

Workflow

  1. Scope one focused role per subagent; pick a scope (.claude/agents/ project vs ~/.claude/agents/ user) → references/authoring-workflow.md
  2. Write a specific description (add "use proactively" where delegation should be eager) and a numbered-workflow prompt body → references/writing-style.md
  3. Restrict tools/disallowedTools to the minimum the role needs; on a coordinator, allowlist spawnable subagent types with tools: Agent(worker, researcher) → references/delegation-and-squads.md
  4. Choose model: (and effort:) using the model-selection skill before finalizing frontmatter — never leave it on the inherit default without a deliberate reason
  5. New agent files in an existing .claude/agents//~/.claude/agents/ load within seconds, no restart — restart only when that scope's agents/ directory didn't exist when the session started
  6. Check references/authoring-workflow.md if a subagent doesn't load or never gets invoked
  7. Check references/undocumented-and-gotchas.md for version-specific behavior (background-by-default execution, model resolution order, output scanning) missing from a quick skim of the docs
  8. Run markdown-best-practices over the finished subagent file — it's still Markdown (heading/list hygiene in the prompt body, no bare fences)

Reference Files

FileLoad When
references/frontmatter-reference.mdWriting or reviewing any frontmatter field, the available-tools list, or file location/priority rules
references/writing-style.mdWriting or reviewing a description or prompt body — delegation phrasing, workflow structure, output format
references/delegation-and-squads.mdBuilding a coordinator + specialist squad, nested subagents, Agent(...) allowlisting, or resuming a subagent
references/authoring-workflow.mdCreating a subagent (via Claude or by hand), choosing its scope, or troubleshooting why it won't load/invoke
references/undocumented-and-gotchas.mdHitting a version-specific behavior the quickstart doesn't cover — background execution, permission-mode inheritance, output scanning, forks vs. subagents

评价

还没有评价。欢迎分享你的使用感受。

评论

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。

提交后会立即公开。请勿填写邮箱、电话等个人信息。URL 最多 2 个。

0 / 1000 字

此表单受 reCAPTCHA 保护,适用 Google 的隐私权政策和服务条款。

同一仓库的技能

Abstract math for 3D rotation representations and conversions. USE FOR: understanding axis-angle representation; Euler angles (yaw/pitch/roll, convention choices); rotation matrices (SO(3)); quaternion-based rotation; converting between representations; identifying and avoiding gimbal lock; composing rotations; half-angle formula; right-hand rule. DO NOT USE FOR: quaternion arithmetic details (use quaternion-algebra); slerp/lerp interpolation (use interpolation-on-manifolds); coordinate system definitions (use coordinate-system-conversions).

中文简介正在准备中。当前显示原文说明。

NetFabric/NetFabric.Numerics362026年8月10日 更新

Design and scaffold multi-agent squads/swarms using the graph orchestration pattern (nodes = agents, edges = dependencies/data flow). USE FOR: designing a coordination topology for a multi-agent squad (sequential pipeline, parallel fan-out/aggregation, branching, feedback loop, cyclic); building an orchestrator agent that delegates to specialist agents; setting up an adversarial/tribunal code or document review with independent reviewers; scaffolding a squad's agent files for a target coding-agent harness (e.g. Copilot CLI or Claude Code custom agents). DO NOT USE FOR: authoring a single agent's frontmatter/prompt body (use copilot-cli-custom-agents, claude-code-custom-agents, or the target harness's own skill); SKILL.md authoring (use create-skill); ad-hoc multi-agent chat with no explicit topology.

中文简介正在准备中。当前显示原文说明。

NetFabric/NetFabric.Numerics362026年8月10日 更新

Find, inspect, query, redact, and troubleshoot logs, transcripts, sessions, and OpenTelemetry data from coding-agent harnesses on Windows, macOS, Linux, and WSL. USE FOR: GitHub Copilot in VS Code; Copilot CLI; Claude Code; OpenAI Codex; Gemini CLI; OpenCode; Cursor, Windsurf, Kiro, or other VS Code-family editors; locating log files; diagnosing tool or model failures; correlating turns, subagents, and tool calls; querying JSONL, plaintext, or SQLite without spending model tokens; preparing sanitized diagnostics. DO NOT USE FOR: application logs unrelated to coding agents; uploading raw logs; treating unstable transcript schemas as public APIs.

中文简介正在准备中。当前显示原文说明。

NetFabric/NetFabric.Numerics362026年8月10日 更新

agents-md

免费

Author and maintain AGENTS.md files — the open, tool-agnostic format for giving coding agents project context. Use when: creating AGENTS.md; writing an agent instructions file; root or nested AGENTS.md for monorepos; migrating CLAUDE.md, AGENT.md, .cursor/rules, or copilot-instructions.md to AGENTS.md; greenfield project setup before code exists; brownfield onboarding for an existing codebase; documenting build/test/lint commands, code style, PR/commit conventions, and security considerations for agents; AGENTS.md discovery and precedence rules. Works for any language, framework, or project size. DO NOT USE FOR: authoring SKILL.md skill files (use create-skill); APM package/plugin authoring (use apm-producer); language-specific style rules (use language skills e.g. csharp-best-practices).

中文简介正在准备中。当前显示原文说明。

NetFabric/NetFabric.Numerics362026年8月10日 更新

Abstract math for angle measurement, unit systems, and circular arithmetic. USE FOR: implementing angle types; converting between degrees/radians/gradians/revolutions; reducing angles to canonical range [0, Full); computing reference angles and quadrants; implementing safe modular arithmetic with negative inputs; angle addition and subtraction; classifying angles by sign and size. DO NOT USE FOR: trigonometric functions (use trigonometric-functions); coordinate conversions (use coordinate-system-conversions); interpolation (use interpolation-on-manifolds).

中文简介正在准备中。当前显示原文说明。

NetFabric/NetFabric.Numerics362026年8月10日 更新

Install, verify, and integrate codebase-memory-mcp (CBM) — a local C binary indexing a codebase into a persistent knowledge graph (functions, classes, call chains, routes) queryable via 15 tools. USE FOR: creating or editing any agent/skill/prompt/subagent that reads, searches, or reasons about a codebase — CBM's CLI must replace grep/ripgrep/glob-and-read exploration for structural questions (who calls X, dead code, architecture overview, diff impact, Cypher queries); checking whether `codebase-memory-mcp` is installed; installing/updating it; checking whether a project's index is fresh/in sync (`index_status`); choosing CLI vs MCP (always prefer the CLI, far fewer tokens per call than MCP); picking the right CBM tool per workflow stage (discovery, verification, implementation, review). DO NOT USE FOR: semantic/business-logic code review; non-code repos with nothing to index.

中文简介正在准备中。当前显示原文说明。

NetFabric/NetFabric.Numerics362026年8月10日 更新

查看 NetFabric 的全部技能

报告此技能的问题