Stack Status
OverKill Hill P³ · overkillhill.com · github.com/OKHP3
A single command that answers "is my local-first AI lab healthy right now,"
without opening five different apps.
What to check
- Ollama:
curl -s http://127.0.0.1:11434/api/ps for currently loaded
models (empty response array means idle, which is normal and good), and
curl -s http://127.0.0.1:11434/api/tags | jq -r '.models[].name' for the
full local model roster count. If the connection is refused, say plainly
that Ollama isn't running or isn't listening on 11434, don't guess why.
- Docker containers:
docker ps --format '{{.Names}}\t{{.Status}}\t{{.Image}}'
filtered to the expected local-AI containers (open-webui, searxng,
qdrant). Flag any of the three that are expected but not running.
- Docker resource ceiling:
docker system info --format '{{.MemTotal}}'
or equivalent, to confirm the VM's memory cap is still in place (should be
a fixed number, not unbounded). This is a known guardrail from an earlier
review; if it's gone, flag it.
- Optional host memory signal: on macOS, if
vm_stat is available,
report compressed pages converted using its reported page size, not a
hard-coded size. Free pages near zero alone do not establish pressure.
On Windows, if PowerShell CIM is available, use
Get-CimInstance Win32_OperatingSystem to report FreePhysicalMemory
and TotalVisibleMemorySize (both KiB). These are capacity figures, not
equivalent to macOS compression or proof of memory pressure. If neither
probe is available, report memory status as unknown and continue the
Ollama and Docker checks. Never gate this portable skill on vm_stat.
Output
A compact status table: component, state, detail. End with one plain-English
verdict: "stack healthy" or a specific named problem plus the two or three
components involved. Don't pad a healthy result with reassurance; a one-line
"all clear" is the correct length when everything's fine.
What this skill never does
Never restarts, stops, or reconfigures a container or the Ollama service.
Report, don't remediate, unless explicitly asked to fix something after
seeing the report.
About
Built by Jamie Hill · OverKill Hill P³
Published at github.com/OKHP3
Part of the OKHP3/skillz Agent Skill library.
MIT License -- free to use, fork, and adapt. A nod to the source is appreciated.
1---2name: okhp3-openclaw-stack-status3description: One-shot health report on the local AI stack -- Ollama, LM Studio, and the Docker-hosted Open WebUI/SearXNG/Qdrant containers. Use when asked "how's my local AI stack", "is Ollama running", "check Docker containers", or before loading a big model.4license: MIT5metadata:6 source: https://skillmd.com/skills/okhp3/okhp3-openclaw-stack-status7 author: Jamie Hill (OverKill Hill P³)8 version: "0.1.1"9 category: openclaw10 origin: okhp3/skillz11 homepage: https://overkillhill.com12 author-github: https://github.com/OKHP313 in_scope: "Report local AI service and container health."14 out_of_scope: "Restarting, stopping, or reconfiguring services without an explicit request."15 openclaw:16 requires:17 bins: [curl, docker, jq]18---1920# Stack Status2122**OverKill Hill P³** · [overkillhill.com](https://overkillhill.com) · [github.com/OKHP3](https://github.com/OKHP3)2324A single command that answers "is my local-first AI lab healthy right now,"25without opening five different apps.2627## What to check28291. **Ollama**: `curl -s http://127.0.0.1:11434/api/ps` for currently loaded30 models (empty response array means idle, which is normal and good), and31 `curl -s http://127.0.0.1:11434/api/tags | jq -r '.models[].name'` for the32 full local model roster count. If the connection is refused, say plainly33 that Ollama isn't running or isn't listening on 11434, don't guess why.342. **Docker containers**: `docker ps --format '{{.Names}}\t{{.Status}}\t{{.Image}}'`35 filtered to the expected local-AI containers (open-webui, searxng,36 qdrant). Flag any of the three that are expected but not running.373. **Docker resource ceiling**: `docker system info --format '{{.MemTotal}}'`38 or equivalent, to confirm the VM's memory cap is still in place (should be39 a fixed number, not unbounded). This is a known guardrail from an earlier40 review; if it's gone, flag it.414. **Optional host memory signal**: on macOS, if `vm_stat` is available,42 report compressed pages converted using its reported page size, not a43 hard-coded size. Free pages near zero alone do not establish pressure.44 On Windows, if PowerShell CIM is available, use45 `Get-CimInstance Win32_OperatingSystem` to report `FreePhysicalMemory`46 and `TotalVisibleMemorySize` (both KiB). These are capacity figures, not47 equivalent to macOS compression or proof of memory pressure. If neither48 probe is available, report memory status as unknown and continue the49 Ollama and Docker checks. Never gate this portable skill on `vm_stat`.5051## Output5253A compact status table: component, state, detail. End with one plain-English54verdict: "stack healthy" or a specific named problem plus the two or three55components involved. Don't pad a healthy result with reassurance; a one-line56"all clear" is the correct length when everything's fine.5758## What this skill never does5960Never restarts, stops, or reconfigures a container or the Ollama service.61Report, don't remediate, unless explicitly asked to fix something after62seeing the report.6364## About6566Built by [Jamie Hill](https://overkillhill.com) · [OverKill Hill P³](https://overkillhill.com)67Published at [github.com/OKHP3](https://github.com/OKHP3)68Part of the [OKHP3/skillz](https://github.com/OKHP3/skillz) Agent Skill library.69MIT License -- free to use, fork, and adapt. A nod to the source is appreciated.