loom-mine-history

Mine git/PR history to surface decisions and provenance from brownfield repositories.

Updated May 3, 2026
One-click install
npx skills add https://github.com/fkberthold/loom --skill loom-mine-history
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: loom-mine-history
Source: https://github.com/fkberthold/loom/tree/main/skills/loom-mine-history
Command: npx skills add https://github.com/fkberthold/loom --skill loom-mine-history

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Mine brownfield repo git/PR history to surface in-flight decisions and file provenance-mined drawers in MemPalace.

Core Features & Use Cases

  • Two-pass cost gate: dry-run preview followed by explicit user go-ahead before the paid pass.
  • Wrapper orchestration + MCP filing: resolves wing location, writes per-unit drafts and arc structures, and updates history_mined_through facts.
  • Incremental mining and archiving: supports resuming from watermark and preserves outputs as drafts.jsonl, kg-triples.jsonl, and arcs.jsonl when synthesis is requested.

Quick Start

Run the dry-run to preview the history mine, then confirm to perform the paid pass and file results.

Frequently Asked Questions about loom-mine-history

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I extract architectural decisions from git history in a legacy codebase?

Mining git history surfaces salient decisions and provenance from a brownfield repository's pull requests, generating per-unit drafts and KG triples for legacy codebases predating structured integration.

What is the best way to capture provenance from pull requests without excessive API costs?

A two-pass cost gate manages expenses by running a dry-run preview first, requiring explicit user go-ahead before executing the paid mining pass and filing provenance results.

Can I resume mining git history if the process is interrupted halfway through?

Yes, incremental mining supports resuming from a maintained history_mined_through watermark, allowing you to continue archiving git and pull request provenance without reprocessing previously mined commits.

How do I file mined repository decisions into structured knowledge graphs?

The mining process orchestrates wrapper invocation and MCP filing to write per-unit drafts and KG triples directly to the wing, preserving outputs as drafts.jsonl and kg-triples.jsonl files.

Does this git history mining approach work for brownfield codebases without prior documentation?

Yes, it specifically applies to legacy codebases predating structured knowledge integration, using commit and pull request history to retrospectively surface and file undocumented architectural decisions and provenance.