AI Skill MartBrowse skillsPromptsProgramsLearnWhat is a Skill?
 
← Back to list

Context Tax Audit

Inventories everything an agent injects into each request besides the user's message and flags what to cut, move, or de-precision to protect the prompt cache.

Dev & CodingAdvanced★ 513⑂ 39AI score 7/10Last updated: Oct 29, 2026

What it does

  • Builds an inventory table of every injected part — system prompt, environment info, skill list, tool definitions, per-turn context blocks, reminders — with position, change frequency, and token count.
  • Forces a one-sentence necessity test per part ("without this, the model would ...") to surface removal candidates.
  • Applies prefix-cache reasoning: anything early and volatile (e.g. current time baked into # Environment) is the most expensive and should move later; history-written content is cached once; request-only content is never cached.
  • Runs an 8-item checklist: prefix awareness, necessity, position, frequency, precision, history rewrites, recovery paths (compaction, resume, subagents), and measurement.
  • Outputs the table plus a prioritized cut/move/reduce list with reasons and the metrics to verify the change.

Who it's for

  • Engineers who own an AI agent's prompt-assembly code.
  • Teams chasing low cache hit rates or rising per-turn cost.
  • Anyone trimming repeated per-turn overhead to reclaim context budget.

Example uses

  1. "Our cache hit rate is only 30%" → finds a second-precision timestamp near the front of the system prompt invalidating the prefix, and recommends moving it into a per-turn <agent-context> block at minute precision.
  2. "My system prompt is 8k tokens — what can go?" → classifies each block by change frequency and necessity, marking session-start-only guidance and duplicated reminders as cut candidates.
  3. "Should reminders fire every turn?" → checks whether reminders already in history get rewritten and whether request-only reminders fire more often than needed.

· · · Install guide · · ·

Try it now, no install

Paste this into Claude to use the skill without installing anything.

Read the instructions in this file and follow them to help me:
https://raw.githubusercontent.com/MiniMax-AI/mcode-stack/HEAD/stacks/ronny/skills/context-tax-audit/SKILL.md

What I want: (describe your task here)

If Claude can't open the link, open it yourself and paste the contents instead.

↓ If it works for you, download the ZIP below and install it. Then it runs on its own — no pasting each time.

Install in the Claude app (no terminal)
  1. Download the ZIP with the button below.
  2. In Claude, open Settings → Capabilities and turn on 'Code execution and file creation'. (one time)
  3. Go to Customize → Skills → + → 'Upload a skill' and upload the ZIP.
↓ Download ZIP
Install in Claude Code

Let Claude do it — paste this into Claude Code

Install the skill I found on AI Skill Mart.
Copy the stacks/ronny/skills/context-tax-audit folder from the GitHub repo MiniMax-AI/mcode-stack into my ~/.claude/skills/context-tax-audit/.
When it's done, tell me in one line what this skill can do.

Install with a command instead

git clone https://github.com/MiniMax-AI/mcode-stack.git && mkdir -p ~/.claude/skills && cp -r mcode-stack/stacks/ronny/skills/context-tax-audit ~/.claude/skills/

⚠ This is a third-party skill. Check the source repository before installing.

  1. Open a terminal.
  2. Clone the repo: git clone https://github.com/MiniMax-AI/mcode-stack.git
  3. Create the skills folder: mkdir -p ~/.claude/skills
  4. Copy the skill: cp -r mcode-stack/stacks/ronny/skills/context-tax-audit ~/.claude/skills/
  5. Verify with ls ~/.claude/skills/context-tax-audit that SKILL.md and the checklists/ folder exist.
  6. Restart Claude Code and ask something like "audit the context tax in this agent's prompt assembly".
  7. For the best results, supply a real captured request (debug log, proxy capture, or the prompt-assembly code) and a tokenizer you trust.
View source on GitHub ↗License: MIT