docs(architecture): Formalize Platform Hook Architecture for Global Consistency

[aheusingfeld]

This Pull Request introduces the conceptual framework necessary to transition SpecForge from a "soft prompt" skill into a deterministic, harness-gated Agentic Application.

As LLMs suffer from "Context Overflow" and "Eager Execution", relying solely on Markdown instructions for strict architectural and routing governance (Global Consistency) is insufficient.

What's included in this PR:

**1. **
A formal architectural document (written in English) that defines four critical hooks designed to be implemented on the execution platform (e.g., as MCP servers, IDE extensions, or agent intercepts):

• AI Validator Hook (Principle Gate): A secondary LLM call to validate proposed plans against the central .
• Pre-Write Policy Enforcer: A hard intercept on file-writing tools to prevent files from being written to unauthorized directories.
• Context-Graph Traceability: A post-creation soft-recommendation hook that ensures newly created specs actively link back to a Knowledge Graph or Zettelkasten ecosystem.
• RPI Enforcers: Formalization of the State Machine, Context Masking (Sub-Agent Spawning), and Plan Fidelity Diff-Checker concepts.

**2. **
A deterministic CI/CD validation script.

• Constraint: It operates entirely relative to the repository root.
• Function: It asserts the existence and basic integrity of the essential SpecForge governance files ( and ), demonstrating how these rules can be statically checked in a GitHub Action or GitLab CI pipeline without depending on any user's local directory structure.

(Co-Authored-By: Craft Agent)