feat: add prompt rewriter skill and optimization category

[rosspeili]

Purpose

This PR implements the Prompt Token Rewriter skill, introducing the new optimization category to the Skillware registry. This skill serves as a deterministic middleware to help agents compress bloated prompts and context histories, significantly reducing token costs while maintaining 100% semantic integrity.

Changes

• New Skill: skills/optimization/prompt_rewriter/
  • skill.py: Heuristic Regex-based compression logic with three aggression levels (low, medium, high).
  • manifest.yaml: Standardized metadata aligned with Skillware parameters and constitution guidelines.
  • instructions.md: Detailed cognitive map instructing the agent on how and when to use the tool.
  • card.json: UI presentation schema for agent dashboards.
• Documentation Updates:
  • Created docs/skills/prompt_rewriter.md (Reference Card).
  • Updated docs/skills/README.md to include the new Optimization category.
  • Added a "Skill Chaining" middleware example to docs/usage/gemini.md.
• Quality Assurance:
  • Added a robust pytest suite in tests/skills/optimization/test_prompt_rewriter.py.
  • Provided an interactive test script in examples/prompt_compression_demo.py.

Verification Results

• pytest suite: 4/4 Passing (Schema, Low-Aggression, High-Aggression, Error-Handling).
• Manual verification via demo: Successfully reduced a sample conversational prompt from 39 tokens to 15 tokens (~61% saving) while retaining all core instructions.

Fixes #21