Add magellan to partner-built — enterprise knowledge discovery

[apardawala]

Summary

• Adds Magellan enterprise knowledge discovery plugin to partner-built/
• Builds structured knowledge graphs from enterprise materials (code, documents, transcripts, diagrams)
• Surfaces contradictions and open questions as primary outputs — not side effects
• Includes 12 legacy language guides for AS400/mainframe systems (RPG, COBOL, CL, DDS, JCL, CICS, Assembler/370, NATURAL/ADABAS, IDMS, Easytrieve, PL/I, REXX)
• Pure markdown + JSON — zero code dependencies, no MCP server required

Use Cases

• Enterprise modernization: Build shared understanding of legacy IT landscapes for architecture teams
• Technical due diligence: Ingest a data room and produce a queryable knowledge base for M&A assessment
• Compliance audits: Extract business rules, trace them to source documents, highlight gaps
• Platform migration: Understand source systems from whatever documentation and code exists

Plugin Structure

partner-built/magellan/
├── .claude-plugin/plugin.json
├── commands/
│   ├── magellan.md          /magellan — run pipeline or show status
│   ├── add.md               /magellan:add — add files or corrections
│   └── ask.md               /magellan:ask — query the knowledge graph
├── skills/
│   ├── file-conventions/    JSON schemas for all KG file types
│   ├── ingestion/           Fact extraction + 12 language guides
│   ├── graph-building/      Facts → entities + relationships
│   ├── contradiction-detection/
│   ├── cross-domain-linking/
│   ├── summarization/
│   ├── querying/
│   ├── pipeline-review/     Quality gates
│   ├── onboarding-guide/
│   ├── diagram-generation/  C4 diagrams (Mermaid + PlantUML)
│   ├── dashboard-generation/
│   ├── design-generation/   DDD specs, business rules, contracts
│   ├── rules-export/        DMN, JSON, CSV, Gherkin
│   └── api-spec-generation/ OpenAPI + AsyncAPI
├── CLAUDE.md
├── CONNECTORS.md
├── LICENSE                  Apache 2.0
└── README.md

Test Plan

• JSON files parse correctly (plugin.json)
• Follows existing partner-built directory structure
• Commands use YAML frontmatter with description field
• Skills use YAML frontmatter with name and description fields
• No code dependencies — pure markdown + JSON
• No .mcp.json required (self-contained, uses Claude's built-in tools)
• Tested fact extraction against real enterprise workspace (AS400 code, QA manuals, meeting transcripts) — compared favorably to production version
• Apache 2.0 license

Links

• Author: Slalom
• License: Apache 2.0

🤖 Generated with Claude Code