answer-depth-control-plus

Built on affaan-m/ECC by @affaan-m (264,688 stars, MIT). All credit for the original idea to them. This fork improves and repackages it; upstream license preserved in UPSTREAM_LICENSE.

A Claude Code skill that lets you pick how short, deep, or detailed an answer should be before Claude replies.

Made for people who want more control over answer length and token use. It works with your current Claude Code flow.

🎯 Why

Some answers are too short to help.

Others are much longer than you need.

Token counts are hard to guess by hand.

This skill gives you four clear depth choices. It makes a rough size estimate, asks what you want, and then answers at that depth.

⚡ Install

Run one command:

mkdir -p ~/.claude/skills/token-budget-advisor && curl -fsSL https://raw.githubusercontent.com/OWNER/answer-depth-control-plus/main/skill/SKILL.md -o ~/.claude/skills/token-budget-advisor/SKILL.md

Replace OWNER with the GitHub account that hosts this repo.

No package install is needed. The skill is one file with zero added tools.

💬 Usage

Ask Claude Code:

How does a database index work? Let me choose the answer depth.

Expected output:

Answer size estimate

Input: about 15 tokens
Task: Medium
Full answer: about 120 to 300 tokens

Choose a depth:

1. Short, 25%: about 170 tokens
2. Balanced, 50%: about 210 tokens
3. Detailed, 75%: about 260 tokens
4. Full, 100%: about 300 tokens

Reply with 1, 2, 3, 4, or a percent.
These are rough estimates, not exact counts.

Reply with a number or percent:

2

Claude then gives a balanced answer. Your choice stays active for later replies in the same chat.

You can also skip the menu:

Give me the short version.

Explain this at 75% depth.

Give me a full answer under 300 words.

🔧 What we changed vs upstream

• Rewritten from Japanese into clearer, more concise English while preserving the four-depth workflow.
• Adds exceptions for exact length limits and urgent safety help, avoiding unnecessary depth prompts.
• Improves token estimation for mixed content by calculating text and code separately and adds practical rounding guidance.
• Softens unsupported precision claims: estimates are now explicitly rough guides rather than “85–90% accurate.”
• Adds safeguards for output limits and unrealistically small estimates, while simplifying the menu and depth descriptions.

This skill also credits the community Token Budget Advisor project by Xabilimon1.

📄 License

MIT licensed. The upstream MIT license is preserved in UPSTREAM_LICENSE.