klaxon
Skill1 starsChanged 6 months agoMITView on GitHub
Fire a loud triple-Submarine alert when Claude genuinely needs the user's attention RIGHT NOW. Use sparingly. Triggers when forward progress is blocked on the user specifically AND consequences exist if they do not see it soon. Examples: about to take an irreversible action and need explicit approval, prod incident discovered mid-task, deploy gate waiting on real-eyes sign-off, long-running operation idled waiting on input for more than two turns. Also fires on `/klaxon` for manual testing. Skip for normal turn-end chimes (the Stop hook covers that) and routine clarifying questions (the Notification hook covers that). Skip entirely in headless `claude -p` contexts where no human is watching.
In this repository3 other files it gives its agents.
- context-mdSkill
- docs-updateSkill
- open-source-skillSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 19
ALL three must be true:
- Distinctive: line 34
Every turn end. That is the Stop chime, automatic.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
klaxon: "I REALLY need your attention"
Triple-ping the macOS Submarine system sound with 1.333s pauses. Every fire is logged with timestamp, parent process, and reason to ~/.claude/logs/audio-hook.log, so the user can audit overuse.
Why this skill exists
Claude Code's default audio is silent. Many users wire a Stop-hook chime (every turn ends), and some wire a Notification-hook chime (when Claude Code is waiting on you). Both fire often enough that they fade into ambient noise. There needs to be a third tier: a sound the user reaches for once an hour at most, that genuinely cuts through.
That third tier is this skill. Triple Submarine is intentionally heavier than a normal alert. If you fire it for a non-emergency the user will tell you, and you should recalibrate.
When to fire (autonomous)
ALL three must be true:
- Forward progress is blocked on the user specifically. Not "I finished a turn." A real human-in-the-loop gate.
- Consequences exist if they do not see it soon. Destructive op queued, incident in flight, deploy gate, security finding, irreversible action pending.
- Tink (Stop) and Sosumi (Notification) would be too quiet for the gravity. If the situation would be served by a normal chime, use the normal chime.
Concrete examples:
- About to push to a production branch and need explicit ship approval. Fire BEFORE asking.
- Discovered live data loss or a running prod incident mid-task.
- A long-running operation has been blocked waiting on user input for more than two turns.
- About to take an irreversible action (force-push, db drop, mass delete) and need real-eyes confirmation.
Skip for:
- Every turn end. That is the Stop chime, automatic.
- "Done with the task." Silence is fine; the diff is the report.
- Routine clarifying questions. Those are Notification territory.
- Headless
claude -pcontexts. No human is watching, this is a no-op.
How to fire
~/.claude/skills/klaxon/scripts/klaxon.sh "short reason for the audit log"The reason string lands in ~/.claude/logs/audio-hook.log under the KLAXON triple-ping header so the user can audit calibration. If you fired it for something that did not deserve three Submarines, expect to be called out and dial back.
Manual trigger
User typing /klaxon should fire ~/.claude/skills/klaxon/scripts/klaxon.sh "manual test". Smoke test for the chain.
Companion hooks (recommended bundle)
This skill is most useful as the loudest tier of a three-sound ladder:
| Tier | Sound | When |
| Quiet | Tink (1x) | Every assistant turn ends. Stop hook. |
| Medium | Sosumi (1x) | Claude Code is waiting on user input. Notification hook. |
| Loud | Submarine (3x, 1.333s pauses) | This skill. |
The companion script at scripts/play.sh (logger + headless suppression) plus the settings-recipe.json hook block wire up the quiet and medium tiers. See README.md in this skill's repo for the full install.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.