/token-budget

LLM costs are dominated by token consumption. This skill finds excessive token waste and missing budgets before deployment.

What this is for

• Unnecessarily long system prompts
• Context larger than max_tokens (truncation risk)
• Repetitive instructions across messages
• Read-only skill. No code changes.

PROTECTION RULE - never ~/.claude/

Read-only skill. Guard required if write mode added later.

What You Must Do When Invoked

During analysis, assign a confidence level to each finding: proven (confirmed by evidence), likely (strong signal, needs review), or suspected (weak signal).

If /token-budget -help or /token-budget -h (without further arguments) is invoked: output the ## Usage section unchanged and stop.

Otherwise follow these steps in order, skipping none.

Step 1 - Help check

If invoked with -help or -h, output the ## Usage section unchanged and stop.

Step 2 - Clarify target

Clarify -ProjectDir. Get confirmation.

Step 3 - Scan

& "<SKILL_DIR>/scripts/token-scan.ps1" -ProjectDir "<path>"

Step 4 - Classification

For each call site:

• Efficient: appropriate sizing for the task
• Wasteful: could be optimized without quality loss
• Critical: context window exceeded, truncation guaranteed

Step 5 - Write report

File token-budget-report.md in current working directory:

1. Summary - call sites by efficiency.
2. Call site table - critical first. Per site: file, line, estimated input tokens, estimated output tokens, max tokens, truncation risk, recommendation.
3. Open questions.

Step 6 - Summarize

State report path, highlight critical sites that will truncate or waste tokens.

Usage

/token-budget               # interactive
/token-budget <dir>         # scan project
/token-budget -help