Skip to content

anasfik/claude-code-prompts-reference

Repository files navigation

Claude Code — Complete AI Prompt Reference

Curated reference of prompt text used in Claude Code, extracted from source and organized into standalone markdown docs.


What this repo contains

This repository documents prompt logic across Claude Code across all major prompt surfaces:

  • System prompt assembly and policy sections
  • Tool prompts (DESCRIPTION / PROMPT and prompt builders)
  • Skill prompts (bundled slash-command workflows)
  • Browser automation prompt guidance
  • Command-specific prompts (/review, /ultraplan)
  • Additional command prompts (/init, /insights)
  • Memory system prompts (selection, extraction, session memory, consolidation)
  • Service-level prompts (compact, summaries, suggestions, search/title)
  • Internal agent prompts (verification/statusline/teammate/agent generation)
  • Permission/classifier prompts (permission explainer, auto-mode classifier surfaces)
  • Interface prompts (output styles, teleport title/branch)

Quick navigation

Tools

Coverage audit

Skills

Command prompt surfaces

Memory prompts

Service prompts

Internal agent prompts

│ └── auto-mode.md

Permission/classifier prompts

Interface prompts

Claude Code — Complete AI Prompt Reference

Curated reference of prompt text used in Claude Code, extracted from source and organized into standalone markdown docs.


What this repo contains

This repository documents prompt logic across Claude Code across all major prompt surfaces:

  • System prompt assembly and policy sections
  • Tool prompts (DESCRIPTION / PROMPT and prompt builders)
  • Skill prompts (bundled slash-command workflows)
  • Browser automation prompt guidance
  • Command-specific prompts (/review, /ultraplan, /init, /insights, auto-mode)
  • Memory system prompts (selection, extraction, session memory, consolidation)
  • Service-level prompts (compact, summaries, suggestions, search/title)
  • Internal agent prompts (verification/statusline/teammate/agent generation)
  • Permission/classifier prompts (permission explainer, auto-mode classifier surfaces)
  • Interface prompts (output styles, teleport title/branch)
  • Source coverage audit for prompt-bearing files

Quick navigation

System prompt

Tools

Skills

Command prompt surfaces

Memory prompts

Service prompts

Internal agent prompts

Permission/classifier prompts

Interface prompts

Other prompt surfaces

Coverage audit


Repository layout

claude-code-system-promtps/
├── README.md
├── .gitignore
│
├── system-prompt/
│   ├── main-system-prompt.md
│   ├── security-instruction.md
│   └── effective-system-prompt-selection.md
│
├── tools/
│   ├── file-tools.md
│   ├── shell-tools.md
│   ├── search-tools.md
│   ├── agent-coordination.md
│   ├── planning-tools.md
│   ├── task-management.md
│   ├── team-coordination.md
│   ├── scheduling.md
│   ├── worktree-tools.md
│   ├── mcp-tools.md
│   └── misc-tools.md
│
├── skills/
│   ├── batch.md
│   ├── simplify.md
│   ├── debug.md
│   ├── stuck.md
│   ├── remember.md
│   ├── skillify.md
│   ├── loop.md
│   ├── schedule.md
│   ├── claude-api.md
│   └── update-config.md
│
├── browser-automation/
│   └── chrome-prompts.md
│
├── companion/
│   └── buddy-prompt.md
│
├── commands/
│   ├── review.md
│   ├── ultraplan.md
│   ├── init-and-insights.md
│   └── auto-mode.md
│
├── memory/
│   └── memory-prompts.md
│
├── services/
│   └── session-and-meta-prompts.md
│
├── internal-agents/
│   └── internal-agent-prompts.md
│
├── permissions/
│   └── permission-and-classifier-prompts.md
│
├── interface/
│   └── output-and-teleport-prompts.md
│
└── audit/
    └── prompt-source-audit.md

Prompt architecture (high-level)

Claude Code prompt behavior is split into three primary layers, plus supporting prompt surfaces:

  1. Global system prompt (src/constants/prompts.ts)
  2. Tool prompts (src/tools/*/prompt.ts)
  3. Skill / command prompts (src/skills/bundled/*.ts, src/commands/*.ts)

Supporting prompt surfaces include memory extraction/selection prompts, compaction/suggestion prompts, classifier prompts, and internal agent prompts.


Coverage summary

  • Core system prompt and security policy
  • Effective system-prompt selection logic (buildEffectiveSystemPrompt)
  • Full tool prompt set used in the main runtime
  • Bundled skills and command entry prompts (including /update-config)
  • Memory subsystem prompts (selection, extraction, session notes, consolidation)
  • Background/service prompts (compact, away-summary, tool-use-summary, suggestions, semantic search, title generation)
  • Internal agent prompts and teammate addenda
  • Permission/classifier prompt surfaces
  • UI/interface prompt surfaces (output styles, teleport)
  • Source-to-doc audit mapping for prompt-bearing files

How to use this reference

  1. Start in system-prompt/ for global behavior.
  2. Use tools/ for callable tool instructions.
  3. Use skills/ and commands/ for workflow prompts.
  4. Use memory/ and services/ for background model prompt behavior.
  5. Use internal-agents/, permissions/, and interface/ for internal/runtime-specific prompt surfaces.
  6. Use audit/ to verify source-to-doc coverage.

Scope note

  • This reference tracks prompt material from Claude Code source files.
  • No external repository is used as a source of truth for this folder.
  • If an external URL/name appears in quoted prompt text, it is preserved only because it exists in source.

About

Leaked Claude code research material (how it does everything)

Topics

Resources

Stars

Watchers

Forks

Contributors