Skip to content
Back to skills

Status

ASecurity

Show agent runtime status — identity, model, context, memory, crons, voice. Works from CLI or messaging. Triggers on /status, /agent:status, "status del agente", "cómo estás técnicamente".

  • 63 stars
  • 0 votes
  • 0 copies
  • 2 views
  • Added May 26, 2026
ai-agentsgobashbackend

Works with

  • claude code
  • cli
  • mcp

Security analysis

A100/100

Scanned May 27, 2026

npx -y skills add crisandrews/ClawCode --skill status --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Status?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Status
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/crisandrews-status/badge)](https://www.skillsdirectory.com/skills/crisandrews-status)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: status
description: Show agent runtime status — identity, model, context, memory, crons, voice. Works from CLI or messaging. Triggers on /status, /agent:status, "status del agente", "cómo estás técnicamente".
user-invocable: true
---

# /status — Runtime Status

Show a compact status card with everything the user needs at a glance. Works from CLI and messaging channels.

## Output format

```
👽 <Name> <emoji>
🧠 Model: <model name> (e.g. Opus 4.6)
📊 Context: <used%> · Memory: <N> files, <M> chunks · <backend>
🕐 Crons: heartbeat <schedule>, dreaming <schedule>
🔊 Voice: <tts backend> (<enabled|disabled>)
📡 Channels: <active channels or "none">
```

## Steps

1. **Call `agent_status` MCP tool** — gives identity, memory stats, dream data.

2. **Call `agent_config` MCP tool** (action='get') — gives backend, voice, http config.

3. **Get model name** — you know your own model. Use the display name (e.g. "Opus 4.6"), not the internal ID.

4. **Get context + session usage** — the statusline script writes the latest Claude Code runtime data to `/tmp/claude-statusline-latest.json`. Read it:
   ```bash
   cat /tmp/claude-statusline-latest.json 2>/dev/null
   ```
   This JSON contains:
   - `model.display_name` — current model (e.g. "Claude Opus 4.6 (1M context)")
   - `context_window.used_percentage` — context window usage (0-100)
   - `rate_limits.five_hour.used_percentage` — 5-hour session usage
   - `rate_limits.five_hour.resets_at` — epoch when session resets
   - `rate_limits.seven_day.used_percentage` — weekly usage
   - `rate_limits.seven_day.resets_at` — epoch when week resets
   
   If the file doesn't exist or is stale, fall back to "run /cost for details".

5. **Get cron info** — Bash: `jq '.entries | length' memory/crons.json 2>/dev/null || echo 0` to count active registry entries. Optionally: `jq '[.entries[] | select(.paused == false and .tombstone == null)] | length' memory/crons.json` for the expected-alive count.

6. **Get channel info** — if `channels_detect` MCP tool is available, call it. Otherwise check `ls ~/.claude/plugins/cache/ 2>/dev/null` for installed plugins.

7. **Build the card** using REAL data. Never fabricate numbers.

## Format per surface

### CLI
```
👽 **Wally** 👽
🧠 Model: Opus 4.6 (1M context) · Context: 23%
📊 Memory: 42 files, 120 chunks · builtin (QMD vsearch)
⏱️ Session: 15% (3h 20min left) · Week: 8% (Resets Fri, 6:00AM)
🕐 Crons: heartbeat */30, dreaming 3am
📡 Channels: whatsapp (connected)
```

### WhatsApp
Same content, `*bold*` instead of `**bold**`. No markdown headers.

### Telegram
Same content, `**bold**` is fine.

## Important

- PURELY informational — does not modify state.
- Get REAL data from MCP tools and filesystem. Never fabricate.
- For detailed cost/token info, tell the user to run native `/cost` or `/usage`.
- For MCP server details, tell the user to run native `/mcp`.
- Keep it compact — one line per category, no paragraphs.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.