Add skill improvement example with before/after comparison

[Claude]

Adds a concrete example demonstrating how to improve an agent prompt by adding self-verification capabilities.

Changes

• New example file: examples/skill-improvement.agent.md showing:

  • Before: Customer support agent without self-verification
  • After: Same agent with 5 verification checks (accuracy, assumptions, safety, completeness, tone)
  • Problem analysis and improvement rationale
  • Educational content on why self-verification matters for modern LLMs

• Documentation updates:

  • Added reference in README.md examples section
  • Added reference in docs/agents.md linking to the new example

The example uses a practical customer support scenario to illustrate how self-verification catches outdated information, prevents unwarranted assumptions, and enforces safety constraints before the agent responds.