Turn buried notes, old exports, and forgotten files into knowledge you can actually use.
You have years of notes scattered across apps. Old Evernote exports. A Notion dump from a previous job. Notes on your phone you haven't looked at in months. Code from a project you abandoned. Research you bookmarked and forgot.
Most of that knowledge is buried. /mine digs it out — point it at any source, get back a single condensed capsule ready to hand to another agent, read yourself, or feed into a project.
# Claude Code CLI or Desktop
claude mcp add-skill https://github.com/hurleywgly/mineOr manually place SKILL.md at ~/.claude/skills/mine/SKILL.md.
Just describe what you want and where to look:
/mine landing page tips from my Education notes
/mine everything useful in ~/Downloads/notion-export.zip
/mine what does this repo do ~/my_projects/some-codebase
/mine all the frameworks and mental models in ~/notes/
/mine summarize https://docs.example.com
The agent figures out the source type, scope, and strategy. You get back one file — a capsule with scored, themed findings.
| Source | How to reference it | Platform |
|---|---|---|
| iCloud Notes | "my notes", "Notes app", a folder name | macOS only |
| Local files | Any directory path | CLI / Desktop |
| Git repos | Path to a repo, or "this codebase" | CLI / Desktop |
| ZIP archives | Path to a .zip file |
CLI / Desktop |
| Websites | Any URL | All environments |
- Parse — reads your natural language request, determines source type and scope
- Enumerate — lists what's available, shows you the scale, confirms before mining
- Mine — cycles through content in batches, extracts findings, scores each one
- Package — organizes findings by theme into a single capsule file
Every finding is scored on three criteria:
| Criterion | What it measures |
|---|---|
| Transferable | Is it useful beyond its original context? |
| Specific | Does it contain concrete details — not just vibes? |
| Buried | Is it actually hard to find right now? |
Score 2 or 3 out of 3: keep. Below 2: discard.
A single markdown capsule file with:
- Findings grouped by theme
- Each finding scored and categorized (insight, framework, decision, idea, pattern, quote, action, contact)
- Source attribution for every finding
- A confidence rating (high / moderate / low based on how many findings scored 3/3)
- Coverage notes on what was and wasn't reached
For large sources (hundreds of notes, big repos), the agent automatically sets up progress tracking. Run /mine resume to pick up where you left off. Run /mine status to check progress.
- Claude Code (CLI or Desktop app)
- Standard Claude Code tools: Read, Write, Bash, Glob, Grep, Agent
- macOS required for iCloud Notes (other sources work on any platform)
MIT