docs: Add AgentKit platform reboot foundational documentation

[yu-iskw]

This commit establishes the foundational documentation for the AgentKit
platform reboot, transforming coding-agent-fabric into a comprehensive
resource platform inspired by SkillKit.

Phase 0: Naming and Branding Strategy

• NAMING_DECISION.md: Comprehensive naming strategy addressing:
  • npm scope decision (@agentkit)
  • GitHub organization structure
  • Collision mitigation with existing "AgentKit" projects
  • Migration plan from @coding-agent-fabric
  • Domain strategy and SEO considerations

Phase 1: Product Definition and Architecture

Product Specification (AGENTKIT_V1_PRODUCT_SPEC.md)

Complete v1.0 product specification defining:

• Core value proposition: "Write once. Deploy to every agent."
• Product principles (canonical-first, deterministic, secure-by-default)
• Resource model (Skills, Subagents, Rules, Workflows, Context Packs)
• Resource lifecycle (discover → evaluate → parse → validate → translate → install → test)
• Quality assurance (4-dimension scoring: structure, clarity, specificity, advanced)
• Team collaboration (manifest + lockfile system)
• Runtime discovery (REST + MCP APIs)
• User personas and UX goals
• Platform architecture and package structure
• Success metrics and roadmap

Canonical Schemas (docs/schemas/)

JSON Schema definitions for all resource types:

• skill.schema.json: Reusable instruction modules
• subagent.schema.json: Specialized agent definitions
• agent-profile.schema.json: Agent-specific translation rules
• manifest.schema.json: Project manifest (agentkit.yaml)
• lockfile.schema.json: Resolved state (agentkit.lock)
• README.md: Schema documentation and usage guide

All schemas follow JSON Schema Draft 2020-12 with:

• Complete validation rules
• Quality scoring support
• Provenance tracking
• Dependency management
• Extensibility via metadata fields

Translation Model (TRANSLATION_MODEL.md)

Comprehensive translation engine specification:

• Translation pipeline architecture
• Format mapping strategies (Markdown, JSON, YAML)
• Path mapping rules per agent
• Frontmatter handling strategies (YAML, JSON sidecar, strip, HTML comment)
• Agent profile system
• Feature support matrix
• Translation examples (Skills, Subagents)
• Degradation strategies for unsupported features
• Reverse translation (import) capabilities
• Testing framework for translations

Key Differentiators from SkillKit

While inspired by SkillKit, AgentKit extends the concept:

• Broader scope: All resource types, not just skills
• Plugin system: Extensible architecture
• Lockfile: Reproducible environments
• Comprehensive agent support: 6+ agents with clear extensibility
• Quality gates: Enforceable in CI/CD

Next Steps

Phase 2 will implement:

• Quality scoring system in @agentkit/core
• Translation engine with agent adapters
• CLI command structure updates
• Package renaming from @coding-agent-fabric to @agentkit

https://claude.ai/code/session_01S46YuNpH9iFcToTbrZCZUD