agentleFS
shino369shino369/claude-code-personal-workspace/CLAUDE.md

claude-code-personal-workspace

CLAUDE.md4 starsChanged 28 days agoISCView on GitHub

This is a personal workspace for Claude Code to assist with various daily tasks including programming, documentation, research, consultation, translation, and generative-AI development. Package Manager: This workspace uses pnpm 12.14.0, pinned in package.json via packageManager (Corepack). Always use pnpm commands instead of npm; run corepack enable or prefix commands with corepack pnpm if pnpm is not on your PATH. Why Context Management Matters: Quality degrades at 20-40% usage, long before a window is full. The actual window depends on the…

  • Installs packages

The file

We split this CLAUDE.md into the rules it gives an agent

11 rules. About 0 are common, 4 are distinctive, and 7 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 7

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Use external memory: - Write plans and progress into the task directory: tasks/<task-name-yymmdd>/plan.md and NOTES.md.

    line 19
  2. Distinctive:

    The copy-paste reset: - When context gets bloated:

    line 21
  3. Very specific:

    Run /compact to summarize

    line 23
  4. Very specific:

    Run /clear to wipe context

    line 24
  5. Very specific:

    Know when to clear: - If explaining the same thing 3+ times, /clear and reframe.

    line 27
  6. Distinctive:

    Run these before finishing a change:

    line 44
  7. Very specific:

    Ask for a translation, or invoke the trilingual-translator agent directly.

    line 57
  8. Very specific:

    Use WebFetch for simple pages.

    line 61
  9. Distinctive:

    Verify a third-party skill's runtime before invoking it.

    line 65
  10. Very specific:

    Do not use pnpm add -g: its global bin directory is not on PATH, and global installs bypass the workspace lockfile and supply-chain …

    line 74
  11. Distinctive:

    If a runtime is missing, tell the user what to install rather than silently switching tools.

    line 82

≈ “Common” is an estimate: at least 1 in 5 of the 26,980 CLAUDE.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
CLAUDE.md

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

Repository Purpose

This is a personal workspace for Claude Code to assist with various daily tasks including programming, documentation, research, consultation, translation, and generative-AI development.

Package Manager: This workspace uses pnpm 12.14.0, pinned in package.json via packageManager (Corepack). Always use pnpm commands instead of npm; run corepack enable or prefix commands with corepack pnpm if pnpm is not on your PATH.

Context Management

Why Context Management Matters: Quality degrades at 20-40% usage, long before a window is full. The actual window depends on the model (1M tokens by default; set CLAUDE_CONTEXT_WINDOW to override). A UserPromptSubmit hook (.claude/hooks/context_monitor.js) measures real usage from the transcript and injects [Context Monitor] warnings at 20/40/60%, so act on those warnings rather than estimating from message counts.

Best Practices:

  1. Scope conversations per feature - One conversation per task/feature prevents context bleeding. Don't build auth and refactor database in the same session.
  2. Use external memory - Write plans and progress into the task directory: tasks/<task-name-yymmdd>/plan.md and NOTES.md. These persist across sessions, allowing fresh context with preserved knowledge.
  3. The copy-paste reset - When context gets bloated:
    • Copy important info from terminal
    • Run /compact to summarize
    • Run /clear to wipe context
    • Paste back only what matters
  4. Know when to clear - If explaining the same thing 3+ times, /clear and reframe. Fresh context > fighting through degraded context.
  5. Write plans to a file - For complex tasks, outline the approach in tasks/<task-name-yymmdd>/plan.md. It survives context resets and keeps the implementation focused.

External Memory Files:

  • tasks/<task-name-yymmdd>/plan.md - Implementation plan and design decisions
  • tasks/<task-name-yymmdd>/NOTES.md - Progress, findings, and handoff notes

Model Selection: important! use inherit model from the main session

Task Organization

Conventions for creating outputs (task directories under tasks/, content subdirectories, file naming) live in .claude/rules/task-organization.md.

Quality Checks

Run these before finishing a change:

pnpm lint
pnpm typecheck
pnpm test:coverage
pnpm format:check

Available Systems

Translation

Ask for a translation, or invoke the trilingual-translator agent directly. The translation-expertise skill supplies the methodology; engineering-terminology and document-writing add domain and style guidance.

Web Content

Use WebFetch for simple pages. For browser work, use the web-browsing skill and .claude/rules/browser-tools.md: pnpm exec agent-browser ... is the default for ad-hoc interaction, browser-use is for the user's real Chrome or autonomous/cloud work, and Playwright is e2e-only (e2e/*.e2e.*, pnpm test:e2e).

Skill Runtimes

Verify a third-party skill's runtime before invoking it. Full install notes are in README.md under "Skill runtime prerequisites".

  • agent-browser — devDependency; always invoke as pnpm exec agent-browser ... (no global shim is installed)
  • browser-use — browser-use CLI from the browser-use package (uv tool install --python 3.12 browser-use); it depends on browser-harness for troubleshooting
  • playwright-cli — local Playwright (pnpm exec playwright cli ...)
  • skill-creator — Python scripts via uv run --with pyyaml --python 3.12 python ...
  • gh — GitHub CLI for Dependabot PR review and PR attachments; not installed here (https://cli.github.com/)

Do not use pnpm add -g: its global bin directory is not on PATH, and global installs bypass the workspace lockfile and supply-chain settings.

Third-party skills are not committed. Install them with pnpm skills:install (use --skip <name> to omit optional skills, or --only <name> to select specific ones); third-party installs land in .claude/skills/ and remain untracked. After installing or updating them, run pnpm skills:check to verify every lock entry is installed with a matching frontmatter name.

If a runtime is missing, tell the user what to install rather than silently switching tools.

Additional Documentation

Task Memory:

  • tasks/<task-name-yymmdd>/plan.md - Implementation plan and design decisions
  • tasks/<task-name-yymmdd>/NOTES.md - Progress, findings, and handoff notes

Quick Reference:

  • Context degradation starts at 20-40%, not 100% - manage proactively
  • Record complex plans in the task's plan.md so they survive context resets
  • Task files under tasks/ persist across sessions
  • /clear without hesitation - fresh context beats degraded context

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.