24 skills · category

Claude Code Mastery skills for Claude Code

Skills about Claude Code itself: authoring new skills and tuning their descriptions, writing lean CLAUDE.md files, automating with hooks, designing subagents, building MCP servers and plugins, running headless in CI, and keeping long sessions effective. Install these first if you want the agent to configure and extend itself well.

Download all 24 (.zip) Browse the list

Every claude code mastery skill

Each card opens the full playbook, example prompts and install commands.

★ No. 001

Agent Eval Harness

Prove a skill or prompt change helped, run after run

Instead of eyeballing one run, Claude writes a small labeled suite of positive, negative and edge cases from real usage, grades each with deterministic checks on outcome and process before reaching for a judge model, runs every case several times through claude -p or claude plugin eval, and compares per case against a committed baseline. A stdlib runner script handles fixtures, tool-call grading and regression exit codes.

agent-eval-harness 164 lines · 2 files

No. 002

Agent SDK Builder

Ship a custom agent on Claude Code's own loop

Builds agents on the Claude Agent SDK that are safe to run inside a product. Claude starts with a read-only loop that logs every message type, adds domain tools as in-process MCP servers, gates risky calls with canUseTool or hooks, chooses the system prompt preset and settingSources on purpose so CLAUDE.md and skills load only when intended, and bounds every run by turns, time and logged cost.

agent-sdk-builder 232 lines

No. 003

Command to Skill Migration

Move slash commands into skills without breaking a name

Claude inventories your legacy command files, dry-runs a converter that flags name collisions, unknown keys, side effects and positional arguments, then gives each new skill a deliberate invocation setting and real description. Long bodies move into references and scripts, every skill is invoked once with real arguments, and only then are the old files removed in the same commit.

command-to-skill-migration 176 lines · 2 files

No. 004

Context Window Hygiene

Keep long sessions sharp instead of slowly forgetful

Treats the context window as working memory that degrades when stuffed with dead ends and raw logs. Claude checks /context, trims fixed costs like unused MCP servers, scopes every read and command output, keeps the plan and key facts in a file, and picks the right reset: clear for a new task, compact with keep and drop instructions, rewind after a wrong turn, or a fresh session when corrections stop sticking.

context-window-hygiene 161 lines

No. 005

Devcontainer Sandboxing

Skip prompts only inside a boundary you have tested

Claude matches the isolation to the risk: the built-in Bash sandbox for fewer prompts on a laptop, a dev container with a non-root user, persisted config and a default-deny egress firewall for teams, or a disposable container for unattended bypass runs. Every boundary is proven with escape tests, credentials stay scoped, and bypass mode is used only where the tests passed.

devcontainer-sandboxing 197 lines

No. 006

Git Worktree Parallelism

Several Claude sessions, one repo, zero collisions

Lets you run two to four Claude Code sessions at once without them trampling each other. Claude splits work into tasks with disjoint file scopes, creates a worktree and branch per task, makes each checkout runnable with its own env file, port and database, hands every session a written contract, then rebases and merges in a planned order and removes everything afterwards only with your confirmation.

git-worktree-parallelism 186 lines

No. 007

Headless CI Agent

Claude in your pipeline, bounded and injection-aware

Sets up Claude Code to run unattended in scripts and CI without hanging, overspending, or obeying instructions hidden in a pull request. Claude splits privileges so the agent only reads and writes workspace files while plain workflow steps hold tokens, commits the prompt as a reviewed file, validates the JSON result with jq, caps turns and time, and tests the job against an injection fixture.

headless-ci-agent 196 lines

★ No. 008

Hooks Automation

Turn 'always do X' into behavior that never skips

Converts rules you keep repeating into hooks the harness runs every time. Claude picks the right event and matcher, writes a fast script that reads the hook JSON from stdin, tests the allow and block cases by piping sample payloads before registering anything, and explains each block so Claude can recover. Includes ready recipes for formatting, Bash guards, protected paths, test gates and notifications.

hooks-automation 185 lines · 2 files

No. 009

MCP Server Builder

Connect Claude to your systems with tools it picks right

Builds Model Context Protocol servers that Claude can actually use well. Instead of mirroring every REST endpoint, Claude designs a few task-level tools with bounded schemas, compact outputs and fixable error messages, wires stdio or Streamable HTTP with the official SDKs, tests each tool in the Inspector and a scripted client, and registers the server for you or your team without leaking secrets.

mcp-server-builder 235 lines · 2 files

No. 010

MCP Troubleshooting

From 'Failed to connect' to a named cause in six rungs

Claude stops restarting servers at random and works a fixed ladder: claude mcp list and get, running the server command by hand with an initialize probe, the MCP Inspector, then filtered debug logs. It checks whole-entry scope precedence, env expansion, auth state and timeouts, and only calls it fixed after one real tool call succeeds in a fresh session.

mcp-troubleshooting 180 lines

No. 011

CLAUDE.md Architect

Short memory files that Claude actually follows

Rebuilds CLAUDE.md around one rule: every line is loaded on every task, so it must be a command, a non-obvious convention, or a trap. Claude audits all memory scopes, pulls commands from package scripts and CI instead of guessing, prunes generic advice and linter-enforced style, and moves procedures into skills and must-always rules into hooks. An audit script catches broken imports, stale paths and leaked secrets.

memory-file-architect 179 lines · 2 files

No. 012

Multi-Repo Workflows

Change several repos without breaking any consumer

Claude picks the right setup for the job, from --add-dir for a quick sibling edit to a parent workspace folder whose CLAUDE.md maps every repo, and knows what each setup does not load. Cross-repo changes start with a written plan ordered producer first with a compatibility strategy, every repo gets its own branch and its own test run, and pull requests link to each other with an explicit merge and deploy order.

multi-repo-workflows 163 lines

No. 013

Output Styles

Reshape every reply without touching how code gets written

Claude first checks whether a built-in style or CLAUDE.md, a skill, or a hook is the better fit, then writes a short style file with observable format rules, an exceptions section so errors and risks still surface, and the right keep-coding-instructions choice. Every style is checked with a four-prompt comparison against Default before it is committed.

output-styles 164 lines · 2 files

No. 014

Permissions Policy

Fewer prompts for safe work, hard stops for the rest

Designs a layered permission setup instead of clicking yes all day. Claude classifies the actions your repo really performs, allows the routine read-only and workspace work, asks for anything that leaves the machine, denies secrets and irreversible commands, and commits the shared baseline. It also covers what rules cannot stop, such as reading .env through Bash, and adds hooks or the sandbox there.

permissions-policy 178 lines

No. 015

Plugin Marketplace Publisher

Ship your skills, agents and hooks as one install

Packages a scattered set of skills, subagents, hooks and MCP servers into one versioned plugin and publishes it through a git-hosted marketplace. Claude fixes the layout mistakes that stop components loading, rewrites paths for the plugin cache, validates manifests with a bundled checker, proves the install works in a clean session, and commits the settings that prompt teammates to install it.

plugin-marketplace-publisher 215 lines · 2 files

No. 016

Repo Agent Bootstrap

One pass from fresh clone to agent-ready repository

Claude surveys the stack and CI, runs every build, test and lint command before writing it down, then produces a short CLAUDE.md with a docs map, a committed permission baseline that allows proven commands and denies secrets and production, and a pipe-tested formatter hook. It finishes with a four-question acceptance test in a fresh session and lists anything it could not prove.

repo-agent-bootstrap 157 lines · 2 files

No. 017

Session Handoff Notes

Resume long tasks in minutes, not another hour of digging

Writes the note a fresh session or a teammate needs to pick up a long task immediately. Claude gathers the real state from git and test runs instead of memory, separates verified work from assumptions, records half-finished files with line numbers, failed approaches and traps, asks what to do with uncommitted changes, and proves the note works by having a context-free reader name the next action.

session-handoff-notes 173 lines

★ No. 018

Skill Authoring

Write skills that load when needed and teach something

Turns a repeated prompt or tribal procedure into a working Agent Skill. Claude first runs your real requests without the skill to find the actual gaps, then writes a description built from the words users type, a body limited to procedure, thresholds and traps, and supporting files that load only when needed. A bundled validator checks frontmatter, name rules, body size and every referenced file.

skill-authoring 222 lines · 2 files

No. 019

Skill Description Tuning

Make skills fire on the right prompts, and only those

Treats a skill description as a routing classifier and tests it like one. Claude builds a labeled set of realistic prompts and near-miss negatives, runs each in a fresh headless session to see whether the skill actually fires, diagnoses the miss pattern, and rewrites one variable at a time until recall is high, false positives are zero, and sibling skills still win their own requests.

skill-description-tuning 175 lines · 2 files

★ No. 020

Spec-Driven Development

Requirements, design, tasks, code — each traced to the last

Claude researches in plan mode, then writes requirements, design and a task list as files in the repo and stops for approval at each gate. Implementation goes one task at a time with tests named after requirement IDs, ticked tasks carry commit hashes, deviations update the design before coding continues, and a final trace table proves every requirement has a passing test.

spec-driven-development 199 lines

No. 021

Status Line Scripts

Context, cost and branch at a glance, under 50 ms

Claude builds a status line from the documented stdin contract instead of guessing field names, treats every optional field as possibly absent, caches git calls per session so the bar stays fast, and pipe-tests full, empty and malformed payloads before registering it. Ships a stdlib Python status line with a color-coded context bar, cost and 5-hour limit you can adapt.

statusline-scripts 165 lines · 2 files

No. 022

Subagent Design

Delegate noisy work and get back only what matters

Builds subagents that absorb exploration, logs and test output so the main thread receives a short, structured result. Claude picks the smallest tool set and the right model for the role, writes a description the router actually delegates on, and gives the agent a self-contained prompt with a capped output contract, since it sees no conversation history and only its final message comes back.

subagent-design 185 lines

No. 023

Task Decomposition Plans

Plans with checks, risk first, before a line of code

Stops Claude from diving into a large change and stalling halfway. It researches the code read-only in plan mode, asks only the questions that change the design, then writes a plan file with explicit non-goals, vertical milestones that each leave the system working, runnable acceptance checks, and checkpoints before anything irreversible. Execution follows the file, with statuses and decisions updated as it goes.

task-decomposition-plans 190 lines

No. 024

Usage and Cost Tracking

Find where the tokens went, then cut the top three

Claude measures before advising: a transcript report breaks seven days of usage down by day, model, project, session and skill with cache hit ratio and subagent share, then maps the evidence to concrete drivers such as all-day sessions, Opus defaults, inherited subagent models or bloated MCP context. Fixes come with guardrails like budget caps for scripts and a re-measurement on a comparable week.

usage-cost-tracking 141 lines · 2 files

Choosing Claude Code Mastery skills

Install the skills that match requests you make every week; each costs about 100 tokens of standing context until it loads. If you are not sure where to begin, start with Agent Eval Harness, Hooks Automation and Skill Authoring. For background, read 12 Best Practices for Writing Claude Code Skills, Skills vs Slash Commands, Subagents, Hooks, and MCP and Claude Code Skill Not Triggering? 9 Fixes That Work.

By technology: Model Context Protocol skills Python skills Git skills

Install several

Install several at once

The installer accepts any number of skill names. Add --project to install into the current repository.

curl -fsSL https://claudeskills.page/install.sh | sh -s -- agent-eval-harness agent-sdk-builder command-to-skill-migration context-window-hygiene devcontainer-sandboxing git-worktree-parallelism headless-ci-agent hooks-automation mcp-server-builder mcp-troubleshooting memory-file-architect multi-repo-workflows output-styles permissions-policy plugin-marketplace-publisher repo-agent-bootstrap session-handoff-notes skill-authoring skill-description-tuning spec-driven-development statusline-scripts subagent-design task-decomposition-plans usage-cost-tracking

Claude Code Mastery skills FAQ

How do I install every Claude Code Mastery skill at once?

Download the bundle at https://claudeskills.page/downloads/category/claude-code.zip and unzip it into ~/.claude/skills/ (or .claude/skills/ inside a repository). To pick a few, pass several names to the installer: curl -fsSL https://claudeskills.page/install.sh | sh -s -- agent-eval-harness agent-sdk-builder command-to-skill-migration.

Will installing 24 skills slow Claude Code down?

No. Claude Code keeps only each skill's short description in context — roughly 100 tokens per skill — and loads the full instructions only when a request matches. Install the whole category if it fits your work; unused skills cost almost nothing.

Which Claude Code Mastery skill should I start with?

Start with Agent Eval Harness, Hooks Automation, Skill Authoring. They cover the most common requests in this category; add the rest as your work calls for them.