Add Agentic Debugger skill

[Novera AI (whaojie797-design)]

What problem it solves

AI-generated code frequently "looks right" but fails at runtime, and developers often apply fixes blindly without proof. Agentic Debugger enforces a reproduce-first debugging loop: parse the traceback, generate a minimal pytest reproduction, apply the smallest fix, then verify with a guarded test run — every fix ships with red-before / green-after evidence.

Who uses this workflow

• Developers debugging runtime errors and failing tests with Claude Code
• Teams reviewing AI-generated code that compiles but misbehaves
• Anyone who wants a regression test attached to every bug fix

What is included

• agentic-debugger/SKILL.md — follows the repo template (When to Use / What This Skill Does / How to Use / Example / Tips / Common Use Cases)
• agentic-debugger/scripts/ — three helper scripts (parse_trace, make_repro, run_guard) that run on the Python standard library only, no pip installs; all compile-checked and verified against real samples
• agentic-debugger/references/debug-loop.md — the full loop documentation with anti-patterns

Attribution

Inspired by the reproduce-first debugging discipline from Novera-AI-skills, where the skill is actively maintained by @whaojie797-design.

Checklist

• Skill folder uses lowercase-hyphen naming
• SKILL.md follows the repo template
• README entry added in Development & Code Tools, alphabetical order
• No emojis in the entry; attribution included
• Scripts tested (syntax-checked + real-sample runs)