Configuration
Claude Guardian ships with sane defaults and exposes only three knobs. Everything else is hardcoded to keep the decision surface small.
CLI flags
Section titled “CLI flags”| Flag | Default | Description |
|---|---|---|
--max-log-mb | 200 | Max project log directory size in MB |
--hang-timeout | 300 | Seconds of inactivity before declaring a hang |
--auto-restart | false | Auto-restart on crash/hang |
Automatic aggressive mode
Section titled “Automatic aggressive mode”When disk free space drops below 5 GB, aggressive mode auto-enables:
- Shorter log retention
- Lower size thresholds for rotation
- More aggressive trimming
This is a hardcoded guardrail that cannot be disabled. If disk is critically low, the guardian protects the system first.
Concurrency budget
Section titled “Concurrency budget”The budget system uses deterministic cap transitions:
| Attention level | Max concurrent slots |
|---|---|
| OK | 4 |
| WARN | 2 |
| CRITICAL | 1 |
Leases are time-limited (TTL in seconds) and automatically expire. After risk returns to OK, the cap stays reduced for 60 seconds (hysteresis) before restoring to the base cap. This prevents flapping. The budget prevents multiple heavy operations from dogpiling when the system is already under pressure.
Data locations
Section titled “Data locations”All guardian state lives under ~/.claude-guardian/:
| File | Purpose |
|---|---|
state.json | Current daemon state, attention level, and incident tracking |
budget.json | Concurrency leases and cap |
journal.jsonl | Append-only log of every guardian action |
incidents.jsonl | Incident open/close history |
bundle-*.zip | Doctor diagnostics bundles |
archive/ | Archived log directories |