Claude Code Subagents Pack

Ten subagents your main Claude Code session can hand work to, each with only the tools it needs, plus a linter that tells you why an agent file isn't loading.

What it is

10 ready-to-use Claude Code subagents (https://code.claude.com/docs/en/sub-agents) with tight tool allowlists, plus lint_agents.py, which finds the agent files Claude Code silently skips.

Subagent	Tools	What it does
code-reviewer	Read, Grep, Glob, Bash	Reviews the current diff or named files for bugs, risky changes, missing tests and unclear code
debugger	Read, Grep, Glob, Edit, Bash	Investigates a failing test, error message, stack trace or unexpected behavior, finds the root cause and proposes or applies a minimal fix
dependency-upgrader	Read, Grep, Glob, Edit, Write, Bash, WebFetch (worktree)	Upgrades one dependency or framework version in an isolated git worktree, follows the official upgrade guide, fixes breakages and runs the tests
docs-writer	Read, Grep, Glob, Edit, Write	Writes or updates README files, usage docs, docstrings and changelog entries from the actual code
migration-reviewer	Read, Grep, Glob, Bash	Reviews database migrations and schema changes for locking, data loss, irreversibility and deploy-order risks before they run on production
perf-investigator	Read, Grep, Glob, Bash	Finds the cause of slow endpoints, pages, queries, builds or scripts using measurements, and proposes targeted optimizations
pr-writer	Read, Grep, Glob, Bash	Writes a clear pull request title and description, and a commit message, from the current branch's diff and commits
refactor-planner	Read, Grep, Glob	Plans a refactor or larger change: maps affected code, proposes small safe steps with tests at each step, and flags risks
security-reviewer	Read, Grep, Glob, Bash	Audits code for security vulnerabilities: injection, auth and access control, secrets, SSRF, unsafe file handling, dependency risks
test-writer	Read, Grep, Glob, Edit, Write, Bash	Writes or extends automated tests for a given function, module or bug, using the project's existing test framework and style

Reviewers, planners and the PR writer get no Edit/Write tools and are told not to change files (they keep Bash for git diff and tests; drop Bash from tools for a strictly read-only agent). The upgrader runs in an isolated git worktree. Every prompt tells the agent never to push, deploy or run migrations.

What you get

The screenshot is real output from the kit’s own commands. Tested from a fresh download of this exact zip: ./run_tests.sh ran 8 tests OK on Python 3.13 and 3.10.

Real terminal output from Claude Code Subagents Pack: demo and passing tests

Install (30 seconds)

python3 tools/install_agents.py /path/to/project            # -> .claude/agents/ (commit it to share with your team)
python3 tools/install_agents.py --user                      # -> ~/.claude/agents/ (all your projects)
python3 tools/install_agents.py . --only code-reviewer,test-writer --dry-run

Existing files are skipped unless --force. If .claude/agents/ didn't exist when your session started, restart Claude Code once. Then ask "use the code-reviewer subagent on my changes" or @-mention it.

As a plugin: the folder is also a Claude Code plugin (.claude-plugin/plugin.json, agents in agents/), so you can add it to a plugin marketplace; plugin subagents get scoped names like subagents-pack:code-reviewer (plugin names that start with claude- are reserved by Claude Code, so the plugin is called subagents-pack).

lint_agents.py: why isn't my subagent showing up?

python3 tools/lint_agents.py .claude/agents ~/.claude/agents
python3 tools/lint_agents.py agents --plugin --json

It checks what the docs say makes Claude Code skip a file or misbehave without telling you:

frontmatter that isn't on line 1, YAML that won't parse (an unquoted : inside a description is the classic one), missing name or description, names with : or a leading -;
misspelled fields (Claude Code ignores unknown fields, e.g. maxturns instead of maxTurns) and invalid model, permissionMode, effort, color, memory, isolation, maxTurns, autoCompactWindow values;
tool lists where nothing resolves (the subagent refuses to launch), misspelled tools, tools subagents never get (AskUserQuestion, EnterPlanMode...), disallowedTools specifiers that remove the whole tool;
fields ignored for plugin subagents (permissionMode, mcpServers, hooks, initialPrompt), duplicate names, empty prompts.

Exit code 1 on errors, so it works in CI. Run ./demo.sh for the example and ./run_tests.sh for the 8 tests. Python 3.9+, no dependencies, no network.

Customize

Each agent is one Markdown file: frontmatter plus a system prompt. Common edits: set model: haiku for cheap agents, add memory: project so an agent keeps notes in .claude/agent-memory/, add skills: [...] to preload a skill, or narrow tools further.

Honest limits

Subagent prompts guide the model; they don't enforce anything. Permission rules and hooks are the enforcement layer (see the Claude Code Safety Hooks kit). Field list checked against the Claude Code docs on Nov 9, 2026; new fields may appear.

Format and license

Zip, 23 KB, v1.0, 23 files. You may use and modify this kit for your own work and for clients you serve, including paid client work. Not affiliated with or endorsed by Anthropic.

FAQ

Do I need a paid Claude plan?: You need Claude Code; the files are plain Markdown and work with any plan that runs Claude Code.
Will it overwrite my agents?: No, existing files are skipped unless you pass --force.
Can I use them as a plugin?: Yes, the folder is a Claude Code plugin (agents/ plus .claude-plugin/plugin.json).
How do I get the files?: Right after payment, Stripe’s confirmation page shows your download link.
Refunds?: Email hello@youraitools.co within 14 days if it doesn’t work and we can’t fix it.
Can I pay with crypto?: Yes. Pay the same price in USDC or USDT on Base, Ethereum, Arbitrum, Optimism, Polygon or Solana from the "Pay with crypto" button, with no account or KYC. We check the payment on-chain and show your download links as soon as it confirms.

Get Claude Code Subagents Pack

10 Claude Code subagents (reviewer, security, tests, debugger, migrations, upgrades, docs, PRs, performance, refactors) plus a linter. $19 once.

$19 one-time

Buy and download Pay with crypto

Save with a bundle

Claude Code Subagents Bundle: 3 products for $39 ($57 separately, you save $18).

Pairs well with

DownloadFree

Claude Code Subagents Starter (Free)

3 free Claude Code subagent examples (code reviewer, test writer, debugger) and a linter that shows why a subagent file isn't loading.

View Claude Code Subagents Starter (Free)

Download$19

CSV & Excel MCP Server

Read-only MCP server for CSV, TSV and Excel: describe columns, SQL joins across files and sheets, column stats. No uploads, no dependencies. $19 once.

View CSV & Excel MCP Server