Claude Skill MartBrowse skillsQuick linesLearn by videoTerminal guideWhat is a Skill?
 
← Back to list

Context Engineering

A practical playbook for managing token budgets and context quality in Claude sessions through four operations: Write, Select, Compress, Isolate.

UtilitiesIntermediate★ 2,867⑂ 286AI score 9/10Last updated: Aug 31, 2026

What it does

Tackles the root cause of degrading AI output in long sessions — diluted context — with four concrete operations:

  • Write: persist state to CLAUDE.md, NOTES.md, and .claude/memory/ so it survives compaction and session boundaries.
  • Select: retrieve by precision rank — @file → Grep/Glob → subagent exploration → RAG. Core rule: "focused 300 tokens beats unfocused 113K tokens".
  • Compress: /compact focus:, microcompaction, head+tail reads, tool-result clearing, plus explicit compaction triggers.
  • Isolate: push heavy work to subagents, worktrees (claude -w), /btw, or a fresh session.

It also ships a phase-based budget table (planning <20%, implementation <50%, testing <70%, review <85%), a /clear vs /compact vs subagent decision table, an anti-pattern list, and a ready-to-paste CLAUDE.md rules snippet.

Who it's for

  • Anyone running long Claude Code sessions on large codebases and hitting context limits
  • Developers tired of re-explaining state after every auto-compaction
  • Teams already using subagents or worktrees but lacking clear criteria for when to use them
  • Cost-conscious users who want better prompt-cache hit rates

Example uses

  1. Before a large refactor: write goals and constraints to NOTES.md, and add a PostCompact hook that re-injects .claude/critical-context.md after every compaction.
  2. Debugging test failures: instead of dumping full logs, apply the head+tail 20-line strategy, or delegate the whole test run to a subagent to keep the main context clean.
  3. Switching task domains: after finishing a feature, run /compact focus: auth module changes to keep learnings and drop noise; use /clear when the next task is unrelated.

· · · 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/rohitg00/pro-workflow/HEAD/skills/context-engineering/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 Claude Skill Mart.
Copy the skills/context-engineering folder from the GitHub repo rohitg00/pro-workflow into my ~/.claude/skills/rohitg00-context-engineering/.
When it's done, tell me in one line what this skill can do.

Install with a command instead

git clone https://github.com/rohitg00/pro-workflow.git && mkdir -p ~/.claude/skills && cp -r pro-workflow/skills/context-engineering ~/.claude/skills/

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

  1. Open a terminal and clone the repo: git clone https://github.com/rohitg00/pro-workflow.git
  2. Create the skills folder if needed: mkdir -p ~/.claude/skills
  3. Copy just this skill: cp -r pro-workflow/skills/context-engineering ~/.claude/skills/
  4. Restart Claude Code or start a new session.
  5. Run /context to see your model's real context window, then recalibrate the budget tables in the doc to your setup.
  6. Optionally add export CLAUDE_AUTOCOMPACT_PCT_OVERRIDE=50 to your shell profile and paste the rules snippet at the end of the doc into your project's CLAUDE.md.
View source on GitHub ↗