Epic: framework-agnostic MAP architecture layer and MVP

[rajanbor]

Goal

Evolve MAP from a pattern catalog and context compiler into a deterministic, framework-agnostic AI architecture layer while preserving existing workflows.

Delivery phases

• Record the repository audit in docs/project/CURRENT_STATE.md
• Publish product, architecture, specification, security, and ADR documentation
• Add Pattern Schema v1 with valid and invalid fixtures
• Expose deterministic list, show, search, scan, suggest, graph, and validate workflows
• Add stable machine-readable scan and recommendation contracts
• Connect the pattern graph to registry data and CLI output
• Dogfood MAP against this repository
• Update README, roadmap, changelog, and implementation report

Constraints

• Framework-neutral core; adapters at the edges
• No LLM or network dependency in core analysis
• Deterministic output suitable for people and AI agents
• Backward compatible with existing commands and slash-form pattern IDs
• Every schema has human documentation and fixtures
• CI remains green and security audit reports no known vulnerabilities

This issue tracks the staged implementation; individual pull requests should reference it.

[rajanbor]

Completed through merged PRs #104, #105, #106, #107, and #113.

Delivered: repository audit; product and architecture documentation; five AI architecture ADRs; Pattern Schema v1; scan and recommendation result contracts; deterministic list/search/show/scan/suggest/graph/validate commands; backward-compatible command names; registry snapshot freshness CI; GitHub-only install documentation; dogfood report; changelog; and verified build, lint, typecheck, 146 tests, dependency audit, installer, and CLI smoke checks.

Deferred work is explicitly tracked in #108–#112.