[FEATURE] Billing Transparency: Disclose System Prompt Token Count

[FEATURE] Billing Transparency: Disclose System Prompt Token Count

Summary

As a paying customer, I request visibility into the token count of Claude Code's
system prompt, which consumes my purchased context window without disclosure.

Problem Statement

When Anthropic advertises "200k context" or "1M context":

• An undisclosed system prompt (~10-15k tokens estimated) is injected into every request
• This overhead is billed to the customer as input tokens
• The customer has no visibility into what this overhead contains or its exact size
• /cost and /status do not break down system prompt vs. user content

This is equivalent to:

• Buying a "1 litre" bottle that's 10% packaging material
• Being charged for "200 minutes" of calls where 20 minutes are hold music

Legal Basis (Canadian Customers)

Competition Act (R.S.C., 1985, c. C-34)

• Section 74.01: Misleading representations about price or quantity
• Advertising "200k context" while delivering ~185-190k usable context may constitute

a misleading representation

Quebec Consumer Protection Act (R.S.Q., c. P-40.1)

• Section 219: Prohibition of false/misleading representations
• Section 228: Prohibition of undisclosed charges
• Section 253: Requirement to disclose total cost

PIPEDA Fair Information Principles

• Principle 8 (Openness): Policies must be readily available
• Principle 9 (Individual Access): Right to know what information is used

Requested Solution

1. Immediate: Display system prompt token count in /cost output:

``
System prompt: 12,450 tokens
CLAUDE.md: 2,100 tokens
Tool definitions: 3,200 tokens
Conversation: 45,000 tokens
---
Total input: 62,750 tokens
``

1. Documentation: Publish the base system prompt token count in pricing docs

1. Optional: Provide /system-prompt command to view the actual prompt

(or redacted version showing structure/size)

Related Issues

• #13675 - Show baseline context separately (autoclose pending)
• #22435 - Quota accounting transparency (legal analysis included)
• #13652 - statusLine: total_input_tokens should include system prompt

Why This Matters

Customers paying $0.003-$0.015 per 1k input tokens deserve to know:

• What they're paying for
• How much of their "context window" is actually available
• The exact composition of their charges

This is a basic consumer right, not a feature request.