triage

Workflow failure analysis for the sqa-cloud-deployment-pipeline. The /analyze-workflow-failures skill diagnoses GitHub Actions run failures using logs from Swift object storage.

Setup

This repo is meant to be used inside an LXD container. The source checkouts below are bind-mounted into the container from the host — they are read-only reference material for stack trace lookups and are not modified by the skill.

Required bind mounts:

• ~/sqa-cloud-deployment-pipeline/ — GitHub Actions workflow definitions
• ~/cpe/foundation/ — FCE Python library

Required local directories (created on first run, or create manually):

• outputs/ — analysis reports written here as <UUID>-analysis.md
• tmp/ — scratch space during analysis

MCP server

The skill requires the swift-mcp MCP server for all artifact retrieval. Configure it by creating a .mcp.json file in this directory:

{
  "mcpServers": {
    "swift-mcp": {
      "type": "sse",
      "url": "http://<swift-mcp-host>:8000/sse"
    }
  }
}

The server runs as an SSE service. Replace <swift-mcp-host> with the host running swift-mcp.

If the MCP server is unavailable, the skill will stop immediately — no partial analysis.

Usage

/analyze-workflow-failures <solutions-run-uuid>

or

analyze workflow <UUID>