Documentation Index
Fetch the complete documentation index at: https://auto-sop.com/docs/llms.txt
Use this file to discover all available pages before exploring further.
How auto-sop Works
auto-sop is a Claude Code plugin powered by hooks. It installsPreToolUse, PostToolUse, and Stop hooks that silently capture every Claude Code interaction. A scheduler (launchd on macOS, systemd on Linux) runs the learner hourly to detect patterns and write directives to your CLAUDE.md.
All commands are project-scoped — they operate on the current working directory by default. Use --project <path> to target a different project.
Usage
Global Options
| Option | Description |
|---|---|
-V, --version | Output the version number |
--json | Emit stable JSON output (for scripting) |
-h, --help | Display help for any command |
Project Targeting
Most commands accept--project <path> to operate on a specific project directory instead of the current working directory:
All Commands
| Command | Description |
|---|---|
install | Install hooks, scheduler, and managed CLAUDE.md section |
uninstall | Remove hooks, scheduler, and managed section |
migrate | Migrate from old claude-sop package name |
learn-now | Run the learner immediately |
recap | Show learner recap log entries |
candidates | Manage LLM pattern candidates |
status | Show project install status |
doctor | Run install health checks |
recent | Show recent Claude activity |
show | Show full turn or session details |
stats | Show directive-fire metrics |
errors | Show recent errors from errors.jsonl |
revert | Restore CLAUDE.md from backup |
repair | Fix managed-section drift |
purge | Wipe captures without removing hooks |
pause | Temporarily disable capture + learner |
resume | Re-enable capture + learner |
statusline | Print sop status for terminal prompt |
JSON Output
Pass--json to any command for machine-readable output:

