rushindrasinha/ares-skills/ai-agent-kill-switch
Claude Code and agent skill by rushindrasinha/ares-skills
- community
Checking that the SKILL.md file is still in the repository.
Phrase-triggered emergency stop system for AI agents. Instantly halts all agent automation when a trigger phrase is detected in session logs, or via manual flag. Use when setting up a safety kill switch, emergency stop, or pause-all-automation mechanism for autonomous agents. Survives restarts via LaunchAgent watchdog.
Install
npx vskill@latest install rushindrasinha/ares-skills/ai-agent-kill-switchSKILL.md
AI Agent Kill Switch
Hard stop for autonomous agents. Trigger phrase or manual flag → all outbound halts. Designed to be impossible to accidentally bypass.
Architecture
Components
| Script | Role | |
|---|---|---|
scripts/kill_switch.sh | Activate the kill switch (creates flag, pauses gateway) | |
scripts/kill_switch_watchdog.sh | Scans session logs for the armed phrase | |
scripts/kill_switch_trigger.sh | Runs when flag appears (notifications, gateway pause) | |
scripts/check_flag.sh | Source this at top of any agent script to auto-abort |
Quick Start
1. Activate manually
2. Clear the flag
3. Protect any script
At the top of any script that makes outbound calls:
Phrase-Based Trigger (watchdog)
The watchdog scans agent session JSONL files for an exact armed phrase. When detected, it creates the sentinel flag.
Run via cron every minute:
Or via launchd (macOS) — create ~/Library/LaunchAgents/ai.kill-switch.watchdog.plist with StartInterval=60.
Configuration (env vars)
| Variable | Default | Purpose | |
|---|---|---|---|
KILL_SWITCH_STATE_DIR | $HOME/.ai-kill-switch | Where flag/log/state live | |
KILL_SWITCH_LOGGER_PATTERN | _(none)_ | pkill -f pattern to terminate on trigger | |
ARMED_PHRASE | KILL: AGENT STOP NOW | Watchdog trigger phrase | |
SESSIONS_DIR | $HOME/.openclaw/agents/main/sessions | JSONL logs to scan | |
NOTIFY_TARGET | _(empty)_ | Phone/JID to alert on trigger | |
NOTIFY_CHANNEL | whatsapp | Notification channel |
Why Phrase-Based?
Typing a short kill phrase anywhere the agent is listening (WhatsApp, Twitch chat, a shell session) instantly halts it. No UI, no password, no latency. Designed for the moment you realize something is wrong and you want it stopped now.
Choose an armed phrase that:
- Is unlikely to appear naturally in conversation
- Contains a distinctive marker like
KILL:orSTOP-NOW: - You can type reliably under stress
Full Setup
See references/setup.md (if present) for LaunchAgent plist and recovery procedures.
<!-- A·R·E·S 🛡️ | Mumbai | 2026-04-05 | github.com/rushindrasinha -->
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingai-agent-kill-switch/SKILL.md
Popularity
- 7-day trend score
- 0
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/rushindrasinha/ares-skills/ai-agent-kill-switch)