Add AI agent skill for reverse-engineering OFT specs

[redcatbear]

Problem

Projects adopting OpenFastTrace sometimes have code and user documentation but no traced system requirements or design documentation yet. AI agents can help draft those missing specs, but they need a repeatable workflow that starts from user intent, checks implementation evidence, and records contradictions instead of silently guessing.

Proposed Solution

Add an AI agent skill that guides agents through reverse-engineering missing OFT specs:

• inspect user guides and other user-facing documentation first to infer intent,
• draft doc/system_requirements.md with a uniform OFT structure,
• fill gaps from tests and source code after documenting user-facing intent,
• record contradictions and weak inferences under Open Issues,
• derive an arc42-style design from the code and align dsn items with system requirement scenarios,
• provide reusable templates for requirements and design chapters.

Acceptance Criteria

• The repository contains a reusable skill under skills/openfasttrace-reverse-specs/.
• The skill includes a concise SKILL.md workflow.
• The skill includes templates for doc/system_requirements.md, doc/design.md, and arc42-style doc/design/ chapters.
• The skill includes a checklist for evidence gathering and contradiction handling.
• The skill validates with the standard skill validation script.