How do I install the Voice Attention skill?

Run npx skillmds@latest add webdevtodayjason/voice-attention in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Voice Attention safe to use?

This skill has not completed SkillMD's automated safety review yet. Capability flags: docs only. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Voice Attention?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Voice Attention free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Voice Attention?

webdevtodayjason (@webdevtodayjason) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Voice Attention by webdevtodayjason

Name: Voice Attention
Rating: 4.2 (5 reviews)
Author: webdevtodayjason

Voice Attention

Audibly get Jason's attention via macOS TTS when he may be tabbed away. TWO modes. (1) Automated hooks (already installed in all 4 Claude profiles) speak on the Notification event (Claude needs permission / is waiting for input) and on rapid-repeat Stop fires (loop detector). (2) PROACTIVE — Jason granted Claude standing permission (2026-06-04) to ping him on its own initiative when it genuinely needs him and no hook will fire: a long task finished, blocked on a decision, hit an unresolved error, or about to wait. Use this skill whenever you need Jason's attention and he might not be looking at this window.

File contents

voice-attention — pull Jason back to the right window

Jason tabs across many Claude windows all day. This makes Claude speak the project folder name out loud so he knows which folder needs him.

Engine: macOS say (zero-dependency). Shared script: ~/.config/claude-hooks/voice-attention.py. Folder name is always the live cwd basename (e.g. "subctl", "holace", "callscrub.io"). Mute globally with CLAUDE_VOICE_OFF=1; change voice with CLAUDE_VOICE_NAME (default Samantha).

Mode 1 — automated hooks (no action needed; already live)

Installed in ~/.claude, ~/.claude-jason, ~/.claude-titanium, ~/.claude-semfreak (settings.json → hooks):

Notification → "Hey. . needs your attention / permission / is waiting for you."
Stop (loop detector) → if Stop fires ≥5× within 120s (e.g. a blocking /goal hook re-firing), "Hey. . Claude may be stuck in a loop and needs you." Normal single completions stay silent.

Mode 2 — PROACTIVE (Claude pings on its own — permission granted)

When YOU need Jason and no hook will fire, run:

/usr/bin/python3 ~/.config/claude-hooks/voice-attention.py --say "<short reason>"

It speaks: "Hey. . ."

When to use it (genuine attention-needs only):

A long-running task just finished and the next step is his call.
You're blocked on a decision only he can make.
You hit an error you can't resolve and need input.
You're about to wait/idle on something and want him back.

When NOT to use it: routine progress, things you can decide yourself, or rapidly in succession (don't become noise). One ping per genuine attention-need. Keep the reason short and specific — it's spoken aloud.

Examples

# done with a long build, needs review
/usr/bin/python3 ~/.config/claude-hooks/voice-attention.py --say "the cutover plan is ready for your review"
# blocked on a decision
/usr/bin/python3 ~/.config/claude-hooks/voice-attention.py --say "needs your call on the deploy"

Notes

The proactive command works from any profile/project (script path is profile-independent).
To make this skill discoverable in the other 3 profiles too, copy/symlink this dir into their skills/ (the proactive command already works everywhere regardless).

---name: voice-attentiondescription: Audibly get Jason's attention via macOS TTS when he may be tabbed away. TWO modes. (1) Automated hooks (already installed in all 4 Claude profiles) speak on the Notification event (Claude needs permission / is waiting for input) and on rapid-repeat Stop fires (loop detector). (2) PROACTIVE — Jason granted Claude standing permission (2026-06-04) to ping him on its own initiative when it genuinely needs him and no hook will fire: a long task finished, blocked on a decision, hit an unresolved error, or about to wait. Use this skill whenever you need Jason's attention and he might not be looking at this window.---# voice-attention — pull Jason back to the right windowJason tabs across many Claude windows all day. This makes Claude *speak the projectfolder name out loud* so he knows **which** folder needs him.Engine: macOS `say` (zero-dependency). Shared script: `~/.config/claude-hooks/voice-attention.py`.Folder name is always the live `cwd` basename (e.g. "subctl", "holace", "callscrub.io").Mute globally with `CLAUDE_VOICE_OFF=1`; change voice with `CLAUDE_VOICE_NAME` (default Samantha).## Mode 1 — automated hooks (no action needed; already live)Installed in `~/.claude`, `~/.claude-jason`, `~/.claude-titanium`, `~/.claude-semfreak`(`settings.json` → `hooks`):- **Notification** → "Hey. <folder>. needs your attention / permission / is waiting for you."- **Stop** (loop detector) → if Stop fires ≥5× within 120s (e.g. a blocking `/goal` hook re-firing), "Hey. <folder>. Claude may be stuck in a loop and needs you." Normal single completions stay silent.## Mode 2 — PROACTIVE (Claude pings on its own — permission granted)When YOU need Jason and no hook will fire, run:```bash/usr/bin/python3 ~/.config/claude-hooks/voice-attention.py --say "<short reason>"```It speaks: **"Hey. <current folder>. <your reason>."****When to use it (genuine attention-needs only):**- A long-running task just finished and the next step is his call.- You're blocked on a decision only he can make.- You hit an error you can't resolve and need input.- You're about to wait/idle on something and want him back.**When NOT to use it:** routine progress, things you can decide yourself, or rapidlyin succession (don't become noise). One ping per genuine attention-need. Keep thereason short and specific — it's spoken aloud.## Examples```bash# done with a long build, needs review/usr/bin/python3 ~/.config/claude-hooks/voice-attention.py --say "the cutover plan is ready for your review"# blocked on a decision/usr/bin/python3 ~/.config/claude-hooks/voice-attention.py --say "needs your call on the deploy"```## Notes- The proactive command works from any profile/project (script path is profile-independent).- To make this skill discoverable in the other 3 profiles too, copy/symlink this dir into their `skills/` (the proactive command already works everywhere regardless).

Frequently asked questions