Stats
Actions
Tags
From hydra-action-guard
Show safety configuration, recent blocks and warnings, and classify any command for safety.
How this command is triggered — by the user, by Claude, or both
Slash command
/hydra-action-guard:safetyThe summary Claude sees in its command listing — used to decide when to auto-load this command
When the user runs `/hydra:safety`, show the current safety status.
## Data Source
- Read `${CLAUDE_PLUGIN_ROOT}/state/config.json` for current mode (default: balanced).
- Read `${CLAUDE_PLUGIN_ROOT}/state/audit.jsonl` and filter for `"event":"action_blocked"` and `"event":"action_warned"` entries.
## Output Format
## Rules
1. Show current mode prominently.
2. List recent blocks before warnings.
3. Suggest safe alternatives for each blocked command.
4. Show "No blocked or warned actions" if audit is clean.When the user runs /hydra:safety, show the current safety status.
${CLAUDE_PLUGIN_ROOT}/state/config.json for current mode (default: balanced).${CLAUDE_PLUGIN_ROOT}/state/audit.jsonl and filter for "event":"action_blocked" and "event":"action_warned" entries.## Hydra Action Guard
**Mode:** balanced (block dangerous, warn risky)
**Available modes:** strict | balanced | permissive
### Recent Blocks ([N])
| Time | Command | Category | Reason |
|------|---------|----------|--------|
| 14:23 | rm -rf / | filesystem | Recursive delete from root |
### Recent Warnings ([N])
| Time | Command | Category | Reason |
|------|---------|----------|--------|
| 14:25 | git push --force | git | Force push overwrites history |
### How to Change Mode
Write to state/config.json:
- Strict: {"mode": "strict"} — blocks both "block" and "warn" patterns
- Balanced: {"mode": "balanced"} — blocks dangerous, warns risky
- Permissive: {"mode": "permissive"} — warns only, never blocks
npx claudepluginhub enchanter-ai/hydra --plugin action-guard