improve prompt-engineering-patterns skill description and structure

[fernandezbaptiste]

hey @CodeWithBehnam, thanks for putting together this Claude Code docs mirror with auto-hourly refresh. really like the structured skill library covering 237 skills across ai-ml, backend, devops, security and more. I've just starred it.

ran your prompt-engineering-patterns skill through agent evals and spotted a few quick wins that took it from ~63% to ~90% performance:

• rewrote the frontmatter description with specific actions (design system prompts, structure few-shot examples, enforce structured outputs) + natural trigger terms (system prompt, chain of thought, JSON mode) so the skill activates more reliably

• replaced the generic "When to Use" list and "Core Capabilities" section with a concrete 5-step prompt iteration workflow that gives Claude an actionable sequence to follow

• removed Best Practices, Common Pitfalls, and Success Metrics sections that contained general knowledge Claude already has, cutting ~80 lines of noise

these were easy changes to bring the skill in line with what performs well against Anthropic's best practices. honest disclosure, I work at tessl.io where we build tooling around this. not a pitch, just fixes that were straightforward to make.

if you want to review your other skills, two options: I can open a follow-up PR with a GitHub Action that auto-scores skill.md changes on every PR (no signup, no token needed, runs are fully transparent and the action is pinned + source-inspectable). or if you'd rather do it yourself, spin up Claude Code and run tessl skill review.

happy to answer any questions on the changes.

[CodeWithBehnam]

Thanks for the PR and the kind words about the repo, Bap!

The frontmatter description rewrite is a clear improvement — more specific trigger terms and concrete actions make skill activation more reliable. Happy to take that part.

However, I'm not going to merge the bulk deletions as-is. Here's why:

Removing "Core Capabilities" loses valuable structure. That section serves as an index connecting the 6 topic areas to the 300+ lines of code patterns below. Without it, readers land on a 5-step workflow and then hit a wall of code examples with no overview linking them. The content isn't noise — it's navigation.

"Best Practices" and "Common Pitfalls" aren't redundant. These skills are reference docs for humans too, not just Claude context. And even for LLM consumption, explicit in-context guidance outperforms implicit training knowledge for domain-specific advice like "treat prompts as code with versioning."

The performance claims are unverifiable. "~63% to ~90%" — what eval, what metric, what baseline? Without a reproducible methodology, these numbers don't support the deletions.

What I'd accept:

• The frontmatter description change (lines 1-3) — genuine improvement
• The 5-step "Prompt Iteration Workflow" as an addition, not a replacement for the existing sections

If you'd like to update the PR to keep the existing sections and just add the frontmatter + workflow improvements, I'm happy to merge that.