v1.0.0 — initial release

First public release of the prompt-engineering Agent Skill.

Install

npx skills add k-vaca/prompt-engineering-skill@prompt-engineering

What's in it

A skill that turns a rough request into a professionally engineered prompt. It assumes you may not be an expert in the field you're prompting about, infers your intent, and supplies the wording, structure, and edge-case handling itself.

Built from two deep research passes over the empirical prompting literature and official lab documentation, with claims adversarially verified before inclusion. Notable positions that differ from common guidance:

• Chain-of-thought is task- and model-specific, not a universal improvement
• temperature=0 reduces variance but is not reproducible on hosted APIs
• Few-shot recipes are task-local and don't transfer
• Evidence is tiered — verified findings kept visibly separate from vendor testing

Pinning

Findings are current as of July 2026. Pin this tag if you want stable behavior; the skill itself recommends re-running your evals on any model upgrade.

The attached .skill file is the packaged form; the repository directory is the source of truth.