Enhance README with AI error reduction and prompt rules

[incrediblemohnish5]

Added guidelines for reducing AI errors and prompt engineering rules to improve AI interactions.

[jacobjustinharris]

I am new to GitHub, and this looks great! Managing and making sure that everything can be merged is superb as well.

[federicobartoli]

Thanks for taking time to think about this — appreciate the intent. That said, I don't think this can land in its current form, for a few reasons:

1. Duplicates what the existing skills already do. Each rule here maps onto a structured skill that's already in the repo:

   • "Understand the question" → spec-driven-development
   • "Simple answers" → code-simplification
   • "Only change what you were asked" → already enforced in CLAUDE.md (scope discipline) and the skills themselves
   • "Define good enough" → spec-driven-development (acceptance criteria)
   • "Really check" → test-driven-development, verification gates

   The whole project is "do these things rigorously, with structured workflows." Adding a prose summary to the README weakens that positioning rather than strengthens it.

2. Style mismatch. The README is dense and structured (tables, links, terse sections). This adds 326 lines of conversational prose ("This is the big one", "sounds stupid, but it works"). It reads as a personal blog post rather than project documentation.

3. Project conventions. CLAUDE.md explicitly says: "Never: Add skills that are vague advice instead of actionable processes." This content is general advice — by the project's own rules it shouldn't be added.

4. Out-of-place disclaimer. The "USE WITH CAUTION! I AM NOT RESPONSIBLE..." line doesn't belong in a project README.

Suggest closing this PR. If there's a specific gap you noticed in an existing skill (something the rules above cover but the skills don't), opening an issue against that skill — or a small PR enhancing it — would be more impactful than a parallel doc.