Pilot Shell
A ready harness for Claude Code and Codex: spec-driven development, TDD, memory between sessions, quality gates and human checkpoints
Medium risk
We rate an entry medium when the tool runs code, makes network calls or reads project files. Check what exactly it does before installing.
Why this level
- Changes the agent config and adds sub-agents and rules
- Runs checks and commands as part of the process
Install
Manual install
curl -fsSL https://raw.githubusercontent.com/maxritter/pilot-shell/main/install.sh | bashBack up the agent config before installing.
This is third-party code. Review the repository files before installing.
What it does
The installer places a set of rules, sub-agents, skills and settings tuned for engineering work into the Claude Code and Codex config. It sets up a spec-driven process with task breakdown, tests and quality checks before moving to the next step. Memory between sessions keeps project context, and checkpoints require a human at important points. Settings are merged with yours rather than overwritten, and there is a separate uninstall script.
Who it is for. For developers who want to run serious projects in Claude Code or Codex under a strict process.
Good fit when
- You need a repeatable spec-driven development process
- You want TDD and quality gates inside the agent
- You need project memory between sessions
Not a fit when
- You need a quick one-off edit without a process
- You do not use Claude Code or Codex
Example request
Start a new task with the Pilot Shell process: spec first, then tests, then implementationLimitations
The installer changes the agent config, so back up the config directory and settings file before installing, as the README advises. The underlying agent still calls Anthropic or OpenAI models, whose availability from Russia should be checked separately.
How to disable. Run uninstall.sh from the repository; there are flags to also remove added tools and data.
Security check
- Changes the agent config and adds sub-agents and rules
- Runs checks and commands as part of the process
README in short
The README describes Pilot Shell as a context-engineering harness for Claude Code and Codex. Installation is a single script with version selection, an uninstall script and advice on backing up the config. It lists the delivered parts: rules, sub-agents, a three-way merged settings file, plus autonomy and skill-routing benchmarks. The approach centers on spec-driven development, TDD and quality gates.
FAQ
Will the installer wipe my settings?
Claude settings are three-way merged, but the README advises backing up the config directory and settings file first.
Which agents does it support?
Claude Code and Codex.
Related
A skills library that gives coding agents a development process: brainstorming, planning, TDD, subagents and code review
Skills for real engineers by Matt Pocock
Skills For Real Engineers
Small composable skills for engineering with agents: plan grilling, TDD, bug diagnosis, code review and architecture
GitHub toolkit for spec-driven development: the specify CLI adds agent commands and skills to a project, from principles to implementation
Reference MCP servers
Model Context Protocol servers
Official reference MCP servers: Filesystem, Fetch, Git, Memory, Sequential Thinking, Time and Everything