Instruction architecture

Claude Code system-prompt intelligence

Claude Code combines a product harness, selectable model, scoped CLAUDE.md instructions, auto memory, tools, and client-enforced permissions; the full default system prompt is not established here.

Full prompt unavailable Claude Code instruction architecture · reviewed August 11, 2026

Publication context

Read the provenance first

No full Claude Code system prompt is claimed or republished. The page contains original structure and interpretation grounded in official documentation.

Source tier: Tier A · vendor-published
Confidence: high
Reviewed: Aug 11, 2026
Completeness: partial
Checksum: sha256:a4ac23c988b35…
Raw artifact: No full first-party prompt established

The reader organizes official product, memory, and permission evidence under stable anchors without presenting documentation summaries as raw prompt text.

Product and surface identity

source led map

Source-led map:
Claude Code is documented as an agentic coding product available across terminal, IDE, desktop, and web surfaces. Its model can be selected separately.

Reading note. This page records the product harness and its supported surfaces, not a universal Claude model prompt.

Project instructions and memory

editorial normalization

Editorial normalization:
CLAUDE.md files carry user-, project-, local-, and managed-scope instructions. Auto memory is a separate repository-scoped mechanism, and both are treated as context rather than hard enforcement.

Reading note. The official documentation distinguishes written instructions, auto memory, and enforced hooks or settings.

Tools and permission evaluation

editorial normalization

Editorial normalization:
Claude Code can read, edit, and execute through tools. Permission rules evaluate deny, then ask, then allow; read-only access and mutating actions have different approval postures.

Reading note. Permissions are a client-enforced control layer and should not be collapsed into prompt text alone.

Evidence and provenance

Tier A · vendor-published: Claude Code overview — Vendor-published documentation supports factual product and capability claims.

Tier A · vendor-published: How Claude remembers your project — Vendor-published documentation supports the instruction hierarchy record.

Tier A · vendor-published: Claude Code permissions — Vendor-published documentation supports the permission-boundary record.

Snapshot evidence

What this snapshot includes and leaves out.

Product surface: Claude Code terminal CLI
Model context: Selectable Claude modelClaude Code exposes model selection; this package does not pin one model as universal for the product.
Checksum scope: Three-section source-led documentation map; UTF-8, LF line endings.
Raw mode: The selected official sources do not publish a complete default Claude Code system prompt. No independent capture is substituted.