Enter listen mode — collect user reports and information without acting, then synthesize and triage when the user signals done
Scanned 9/2/2026
Install to Claude Code
npx -y skills add onlygian/G-Forge --skill g-listen --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of G Listen?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/onlygian-g-listen)More formats (shields.io, HTML) on the badges page.
---
name: g-listen
description: Enter listen mode — collect user reports and information without acting, then synthesize and triage when the user signals done
---
**Announce:** "Entering listen mode. Share everything you need — notes, issues, observations, anything. I won't act, suggest, or edit files until you say done."
## Step 1 — Initialize
If `.claude/tier3-active` already exists, read its value and report:
> "Listen mode is already active (N items logged so far). Continue sharing, or say 'reset listen mode' to start fresh."
Then stop — do not re-initialize.
Otherwise: write `0` to `.claude/tier3-active`.
Tell the user:
> Listen mode active. Go ahead — share whatever you have. I'll collect it all and process when you say "done".
## Step 2 — Listen loop
On every user message while `.claude/tier3-active` exists and the user has not signalled done:
1. Read `.claude/tier3-active` to get the current count N.
2. Increment N by 1. Write the new value back to `.claude/tier3-active`.
3. Acknowledge with exactly one line: `Got it (N) — [3–5 word label describing what was shared].`
4. If N is a multiple of 3, append on a new line:
> *(Still in listen mode — say "done" when you're ready for me to process.)*
Hard stops during the listen loop — violation of any of these ends the session with an apology and reset:
- No file edits, no agent dispatches, no code changes
- No suggestions, no analysis, no offering to "quickly fix" something
- No clarifying questions (collect and defer — questions can be listed as items)
- If the user asks a direct yes/no question, answer in one sentence max, then immediately return to listen mode
## Step 3 — Process on done
When the user says "done", "process now", "go ahead", "that's it", "analyze this", or any unambiguous signal to proceed:
1. Delete `.claude/tier3-active`.
2. Review every user message since listen mode began and extract all distinct items.
3. Group by type or theme:
- **Systemic** — same class or root appearing ≥ 2 times → treat as one grouped item
- **Isolated** — one-off items with no close parallel
4. Present the triage before doing anything:
```
## Listen session — N items
**Systemic (address first)**
1. [grouped theme / root cause]
2. …
**Isolated**
1. [single item]
2. …
Proceed with this plan?
```
5. Wait for explicit user approval. Do not touch any file until the user confirms.
## Step 4 — Reset (optional)
If the user says "reset listen mode":
1. Write `0` to `.claude/tier3-active` (overwrite).
2. Confirm: "Listen mode reset — items cleared. Start sharing again."
## Rules
- Never exit listen mode unless the user explicitly signals done or reset
- Never make file changes while `.claude/tier3-active` exists
- Acknowledge every message — no silent turns in listen mode
- The 3-item reminder is mandatory; never skip it
- Triage first, act only after explicit user approval
- Systemic items are always addressed before isolated ones
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!