cbrunner

Opus 5.5 System Prompt Components — Development skill for Claude Code

Development community

Claude Opus 5.5 system prompt leak split into readable components (tools, behavior, agentic rules, memory).

How to install Opus 5.5 System Prompt Components

This entry records only its repository, not the path inside it, so there is no exact command to give. Open cbrunner/opus-5.5-system-prompt-components and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Opus 5.5 System Prompt Components does

Claude Opus 5.5 system prompt leak split into readable components (tools, behavior, agentic rules, memory). Reorganized from a public extract for study.

Alternatives in Development

Agentic Stack — One brain, many harnesses 2.2k ★
StatsPAI — StatsPAI is the first Agent-native Python library for causal inference and applied econometrics — unified API 302 ★
Claude Code Source Code Leak — Claude Code Snapshot for Research 271 ★

README

Claude Opus 5.5 system prompt — split components

Public repo: https://github.com/cbrunner/opus-5.5-system-prompt-components

Credit

**All of the underlying text comes from Pliny the Liberator.**

Pliny extracted and published the full Opus 5.5 session dump on 22 September 2026
Source file: elder-plinius/CL4R1T4S — `ANTHROPIC/CLAUDE-OPUS-5.5.md`
Repo: elder-plinius/CL4R1T4S
X: @elder_plinius

This repo only splits and labels that public extract so it is easier to read. It is not an official Anthropic document, and it is not a new leak.

If you use or cite this split, credit **Pliny the Liberator** for the extraction.

What we split

Pliny's file is ~1.9 MB / ~21.5k lines because it is a full **session dump** (system prefix + user turns + tool results + on-demand skill files). The actual system prompt is only the prefix before the first `--- [user turn] ---`:

**262,652 characters / 976 lines ≈ 65k–70k tokens** with tools, or **~24k tokens** of instruction text without the JSON schemas.

Layout

Path	What it is
`00-system-prefix.md`	Full system slot (tools + instructions)
`01-tool-calling-instructions.md`	How to emit `{antml:function_calls}`
`02-functions-block.md`	Raw blob
`03-deferred-tools-note.md`	Tools that load later
`04-claude-behavior.md`	Persona, products, refusals, tone, cutoff
`05-agentic-behavior.md`	Desktop / agent harness rules
`06-between-agentic-and-memory.md`	Skills, container, device link, model id
`07-user-memory.md`	Memory filesystem + privacy
`08-parallel-tool-call-note.md`	Parallel vs dependent tool calls
`functions/`	One pretty-printed JSON schema per tool
`claude_behavior/`	Subsections of
`agentic_behavior/`	Subsections of
`