# Diagnose Linux Hosts with GASP Metrics

Linux host issues are hard to triage when metrics are scattered across services, logs, and hardware layers. This skill uses GASP output to fetch, compare, and explain system health in clear operational terms.

## Install

```bash
npx skillstore add acceleratedindustries/gasp-diagnostics
```

## Metadata

- Status: approved
- Slug: acceleratedindustries-gasp-diagnostics
- Skillstore revision: r1
- Version status: missing
- Tree hash: 33e2b30c8e126f56d6415494bcf1d7e43d91a13ebecdc5f64a0a48077110ecb7
- Author: AcceleratedIndustries
- GitHub username: AcceleratedIndustries
- License: MIT
- Repository: https://github.com/AcceleratedIndustries/gasp/tree/main/skill/gasp-diagnostics
- Ref: d46e62089a6dfc9e14ea02eebd9b8e2ad2dfe1f4
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: allowed
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: network, filesystem, external\_commands
- Quality score: 50
- Quality tier: warning
- Public page: https://skillstore.pages.dev/skills/acceleratedindustries-gasp-diagnostics
- Manifest: https://skillstore.pages.dev/api/skills/acceleratedindustries-gasp-diagnostics/manifest

## Capabilities

- Fetch GASP metrics from hostnames, IP addresses, or .local endpoints over HTTP.
- Interpret GASP JSON for CPU, memory, disk, network, GPU, services, and logs.
- Compare multiple hosts to identify outliers and shared infrastructure issues.
- Apply workflows for leaks, regressions, container issues, GPU load, disk bottlenecks, and service failures.
- Use infrastructure patterns for workstations, Proxmox hosts, container hosts, and desktop environments.
- Analyze pasted GASP JSON without making a network request.

## Use Cases

- Triage A Slow Workstation: Check whether CPU load, memory pressure, disk I/O, services, or logs explain a slow Linux desktop.
- Compare Infrastructure Nodes: Review several GASP hosts and identify the node that differs from normal cluster behavior.
- Investigate Service Instability: Correlate failed units, recent restarts, log errors, resource pressure, and recent changes.

## Prompt Templates

### Check One Host

```
Check {host} with GASP and tell me whether it is healthy. Focus on CPU, memory, disk, network, and failed services.
```

### Compare Hosts

```
Compare GASP health for {host_a}, {host_b}, and {host_c}. Identify outliers and explain likely causes.
```

### Investigate Regression

```
Investigate why {host} became slower after {change_or_time}. Use GASP trends, recent changes, top processes, logs, and service restarts.
```

### Review Pasted Metrics

```
Analyze this pasted GASP output without fetching anything. Prioritize critical issues, evidence, and safe next actions: {metrics_summary_or_payload}
```

## Limitations

- GASP must already expose metrics from a reachable host or pasted output.
- It does not authenticate to hosts, install GASP, or repair systems automatically.
- Recommendations depend on complete metrics and accurate user context.
- Active fetching can expose host telemetry to the assistant session.

## Best Practices

- Ask for the exact host and permission before fetching live metrics.
- Start with summary health, then verify with supporting metrics.
- Compare current values with baselines before recommending action.

## Anti Patterns

- Do not fetch broad host ranges without explicit approval.
- Do not treat high cached memory as a problem by itself.
- Do not recommend service restarts without impact and timing context.

## Security Audit

- Audited at: 2026-07-07T17:39:39.583\+00:00
- Summary: Most external-command and hidden-file alerts are false positives from Markdown formatting or documented Claude skill installation. The confirmed risks are intentional HTTP fetches of GASP metrics from hostnames or IP addresses, which can expose internal telemetry and support broad host diagnostics. No prompt injection text was found in the reviewed files.

## Stats

- Views: 209
- Downloads: 33
- Favorites: 1
- Popularity score: 0