title: "Claude Code Prompt Collection" description: "Community-authored, illustrative prompt templates for AI coding agents — system prompts, tool prompts, agent delegation, memory management, and multi-agent coordination — checked against Claude Code's official docs." section: "Claude" readTime: "4 min"

Claude Code Prompt Collection

A community-authored collection of prompt templates for building AI coding agents. The prompts were written from scratch to illustrate how production coding agents structure safety rules, route tools, delegate to subagents, and manage memory.

Attribution: Source: repowise-dev/claude-code-prompts — MIT License. Independently authored by the RepoWise team. Not affiliated with Anthropic.

What's Inside

Quick Start

1. Browse the Patterns page to understand how production coding agents structure their prompts.
2. Adapt any template from System Prompts or Tool Prompts for your own agent. To change Claude Code itself, use CLAUDE.md, skills, subagents, and hooks instead (see the Core section); you cannot replace its built-in prompts.
3. Replace {{PLACEHOLDER}} values with your stack, tool names, and risk policy.
4. The Agent Prompts section covers specialized subagent prompts you can use as the body of a custom subagent file (.claude/agents/<name>.md) in Claude Code.

How the Prompts Fit Together

┌─────────────────────────────────────────────┐
│           SYSTEM PROMPT                     │
│  Identity · Safety · Code Style · Tools     │
└───────────────────┬─────────────────────────┘
                    │ governs
        ┌───────────┼───────────┐
        ▼           ▼           ▼
  TOOL PROMPTS  AGENT PROMPTS  MEMORY PROMPTS
  Shell · File  Explorer ·     Summary ·
  Grep · Glob   Architect ·    Session Notes ·
  Web · Agent   Verifier ·     Extraction ·
  Ask · Plan    Docs Guide     Consolidation
        │
        ▼
  UTILITY PROMPTS
  Title · Summary · Recap · Next Action

The system prompt is the behavioral contract that governs all other activity. Tool prompts define how each individual tool should be used. Agent prompts configure specialized subagents. Memory prompts manage context across long sessions. Utility prompts handle small UX details like session titles and away recaps.

Prompt Engineering Methodology

These prompts illustrate common patterns in production coding agents:

• Layered priority — system instructions take precedence over developer instructions, then user instructions, then tool feedback (a design convention; in Claude Code, instructions in CLAUDE.md are context, not enforced rules)
• Intent-based tool routing — each tool has clear rules for when to use it and what NOT to use it for
• Risk-tiered caution — low/medium/high risk tiers with different safeguard requirements
• Adversarial verification — verification means running commands and checking output, not reading code and deciding it "looks correct"
• Minimal context memory — compact memory objects with explicit expiry for stale assumptions

See the Patterns page for deep analysis of each pattern with reusable templates.

Official Guidance

For prompting Claude itself, the primary sources are:

• Claude prompting best practices on platform.claude.com (clear and direct instructions, context and motivation, examples, XML tags, roles, long-context placement, tool use, thinking, agentic systems)
• The Claude Code docs for how it works, memory, subagents, skills, hooks, and the tools reference

Current models named in that guidance include Fable 5.1, Opus 5.5, Sonnet 5.5, and Haiku 5.5. The Patterns page summarizes the parts most relevant to coding agents.

Not affiliated with or endorsed by Anthropic. "Claude Code" is used here to describe the subject matter under nominative fair use.