/aiops-skill:context-fetcher

🗂️ Context Retrieval

Retrieve configuration files, runbooks, and relevant messages from GitHub, Confluence, and Slack to support job investigation and incident analysis — using MCP server integrations.

---

When to Use

💡

Invoke this skill when you want to:

• Find the configuration files for a job or service before or during investigation
• Retrieve runbooks or troubleshooting guides from Confluence
• Review recent code changes related to a failure
• Search Slack for team discussions about a specific job or incident
• Gather documentation context before running root cause analysis

Example invocations:

"Get context for job OCP-Workload-1234567"
"Find the configuration files for the ocp-workload-example catalog item"
"Search for runbooks related to namespace provisioning failures"
"Look up Slack messages about the recent OCP deployment issues"

---

Prerequisites

🐙

GitHub MCP Server Required for GitHub

Configure the github MCP server in your Claude Code settings. Enables mcp__github__search_code and mcp__github__get_file_contents.

📄

Confluence MCP Server

Configure the confluence MCP server for access to runbooks, architecture docs, and incident reports. Enables mcp__atlassian__confluence_search and mcp__atlassian__confluence_get_page.

💬

Slack MCP Server

Configure the Slack MCP server to search channel history for messages and replies related to the investigation topic.

ℹ️

MCP server setup: Add server entries to your Claude Code settings file. After adding new MCP servers, restart Claude Code for them to become available. Check that mcp__github__* tools appear in your available tools list before invoking this skill.

---

6-Step Workflow

1. Identify Search Terms [Claude]

   Extracts relevant keywords from the job name, service name, error messages, or incident details to use as search queries across all sources.

2. Search GitHub [MCP]

   Searches repositories for configuration files, source code, CI/CD definitions, and infrastructure-as-code. Uses mcp__github__search_code and mcp__github__get_file_contents.

   ``` # Example search strategies path:config/*.yaml "job-name" extension:yaml "catalog-item-name" repo:org/agnosticv "workload-name" ```

3. Search Confluence [MCP]

   Finds runbooks, architecture docs, configuration guides, and incident reports using mcp__atlassian__confluence_search. Supports CQL queries for advanced filtering.

   ``` # Example CQL queries space = "Operations" AND text ~ "job-name" label:runbook AND title ~ "namespace-provisioning" ```

4. Search Slack [MCP]

   Retrieves channel history and replies using mcp__slack__slack_get_channel_history to find team discussions, incident updates, and documentation links relevant to the investigation.

5. Synthesize Findings [Claude]

   Organizes all results by source (GitHub, Confluence, Slack) with relevance ratings, file paths, page links, and key excerpts to give a structured investigation brief.

6. Capture Feedback [Claude]

   Calls the feedback-capture skill to record the user's experience with the investigation. This happens automatically at the end of every context-fetcher invocation.

---

GitHub Search Capabilities

The skill searches GitHub for:

• Configuration files — YAML, JSON, TOML (e.g., config.yaml, values.yaml, deployment.yaml)
• Source code — Role tasks, playbooks, Helm templates referenced by the job
• Documentation — README files, docs directories, markdown files
• CI/CD definitions — GitHub Actions workflows, Jenkinsfiles
• Infrastructure as Code — Terraform, Ansible, Helm charts

Search strategies:

Strategy	Example
By file path pattern	path:config/*.yaml "catalog-item"
By content keywords	"job-name" AND "config"
By file extension	extension:yaml "workload"
In specific repository	repo:org/agnosticv "search-term"

---

Confluence Search Capabilities

The skill retrieves from Confluence:

• Runbooks — Operational procedures and troubleshooting guides
• Architecture documentation — System designs and component interactions
• Configuration guides — Setup and configuration instructions
• Incident reports — Historical incidents and resolutions
• Knowledge base articles — Best practices and common issues

Search strategies:

Strategy	Example
By space filter	space = "Operations"
By label	label:runbook, label:troubleshooting
By content	text ~ "namespace-provisioning"
CQL advanced	space = "Eng" AND text ~ "job-name"

---

Configuration

MCP servers are configured in .claude/settings.json or the global Claude Code settings:

{
  "mcpServers": {
    "github": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-github"],
      "env": { "GITHUB_PERSONAL_ACCESS_TOKEN": "<your-token>" }
    },
    "confluence": { },
    "Slack": { }
  }
}

After adding MCP servers, restart Claude Code to activate them.

---

← Back to Skills Next: Feedback Capture →