Demonstrate how to automatically update your Packmind playbook by mining external data sources — using Agent Skills and MCP servers.
AI agent agnostic — Examples below use Claude Code for illustration, but the same skills and MCP-based approach work with any AI coding agent that supports MCP (GitHub Copilot, Cursor, Windsurf, etc.).
flowchart LR
Source["External Source<br>(GitHub, GitLab, Slack,<br>Confluence, ...)"] -->|MCP Server| Skill["Source Skill<br>(fetch & classify)"]
Skill -->|findings report| Update["packmind-update-playbook"]
Update -->|CLI| Packmind["Packmind<br>Change Proposals"]
Each use case follows the same pattern:
- A source skill fetches data via an MCP server and classifies it for playbook relevance
- The classified findings are handed off to packmind-update-playbook, which creates change proposals in Packmind
| Use Case | Source | Link |
|---|---|---|
| GitHub PR Comments | Merged PR review comments | README |
| Slack Conversations | Slack channel discussions | README |
| Jira Issues | Resolved Jira issues | README |
| GitLab MR Comments | Merged MR review comments | README |
| Confluence Documentation | Confluence pages & spaces | README |
| Notion Documentation | Notion pages & databases | README |
The same pattern works with any data source that has an MCP server — Jira, Confluence, Notion, Linear, and more. To add a new source:
- Create a source skill that fetches and classifies data via the relevant MCP server
- Output a structured findings report
- Hand off to
packmind-update-playbook
- A Packmind account and organization
PACKMIND_API_KEY_V3environment variable set- Packmind CLI installed:
npm install -g @packmind/cli - An AI coding agent with MCP support (e.g. Claude Code, GitHub Copilot, Cursor)