Al-Meezan adapter lands alongside a 6.3k-line planning doc dump
Gadoes adds a small MCP adapter for the Al-Meezan Pakistan legal database, but buries it inside a commit that accidentally publishes 6,300 lines of internal planning material - specs, roadmaps, mockups, and Claude skill files. The planning docs are deleted two commits later. The adapter itself is 53 lines.
The functional work is backend/src/lib/mcp/servers/alMeezan.ts (53 lines) plus alMeezan.test.ts (189 lines, 6 tests). It follows the same buildXConfig/X_TOOLS/X_SYSTEM_PROMPT shape established by the CourtListener and GovInfo adapters. The RTL excerpt handling added in Chunk 5 was clearly designed for sources like this - Arabic/English bilingual content.
The rest of commit efa1a03 is noise: .claude/skills/agent-browser/SKILL.md (779 lines), .claude/skills/diagram/SKILL.md, three versions of data/specs/mcp-integration-spec*.md (the main spec is 1,503 lines), data/roadmap/disputes_fork_roadmap.md, a legal_source_picker_v2.html mockup, architecture diagrams, and a repo overview report. Commit e29591e deletes all of it and adds .claude/, data/, and docs/ to .gitignore.
The accidentally-published planning files are not worth importing as artifacts. But if you want to understand where Gadoes is taking this fork - international arbitration focus, multi-jurisdiction sources, citation verification as a hallucination-suppression layer - the deleted data/specs/mcp-integration-spec.md and the disputes roadmap are the most informative single documents in the repo history. Check out efa1a03, read those files, then stay at e29591e.
Spotted something wrong? Or know the PR text has fresher detail than the writeup above?