thinking-beast-mode — Claude Code subagent setup
Autonomous Claude Code subagent designed to drive complex multi-step engineering tasks through verification without yielding control early.
npx claude-code-templates@latest --agent development-tools/thinking-beast-modeWhat thinking-beast-mode does
This subagent tackles long-running engineering workflows that require stubborn persistence. Instead of stopping at a superficial patch, it maps out a visible markdown checklist, digs for underlying root causes across the codebase, edits files incrementally, and immediately verifies changes with Bash commands.
The definition relies on the sonnet model alongside the acceptEdits permission mode. Built-in capabilities include Read, Bash, Grep, Glob, Edit, Write, WebFetch, and WebSearch. When local code patterns fall short, the agent reads online documentation directly rather than relying on outdated internal assumptions.
Destructive operations like force-pushing or rm -rf require explicit operator approval. If an interrupted run is resumed, the agent checks the last open checklist item and picks up the workflow from that exact milestone.
What it helps with
- repair failing test suites until every assertion passes
- implement backend or frontend features end-to-end with tests
- isolate intermittent failures across project files
- verify external library usage against current web documentation
Example prompts for the agent
- Fix our broken test suite after the package upgrade and keep going until every test passes.
- Implement rate limiting for the public API, write integration checks, and confirm it holds up under load.
- Resume the previous run from the last unfinished checklist step.
How to install thinking-beast-mode
1claude-code-templates
npx claude-code-templates@latest --agent development-tools/thinking-beast-mode2Manually: put the files in place
.claude/agents/thinking-beast-mode.mdFor all projects use ~/.claude/agents/. Claude delegates by the description field or an @-mention
Manually: put the files in place
.codex/agents/thinking-beast-mode.tomlCodex defines subagents in TOML: name, description, developer_instructions. Move the Markdown text into developer_instructions by hand. For all projects use ~/.codex/agents/
Manually: put the files in place
.cursor/agents/thinking-beast-mode.mdCursor also reads .claude/agents/
Step by step, with commands for each platform: Claude Code subagents: how to create one, call it and keep context clean
Questions about thinking-beast-mode
How to install thinking-beast-mode in Claude Code?
Run npx claude-code-templates@latest --agent development-tools/thinking-beast-mode in your terminal or place the agent prompt file into .claude/agents/thinking-beast-mode.md.
What does thinking-beast-mode do?
It operates as an autonomous agent that investigates codebases, writes out an actionable todo list, implements fixes, and continuously runs test suites until real verification is achieved.
Which tools does the agent access?
The configuration grants access to Read, Bash, Grep, Glob, Edit, Write, as well as WebFetch and WebSearch for network lookups.
