Add eval-driven development skill

[yiouli]

Pull Request Checklist

• I have read and followed the CONTRIBUTING.md guidelines.
• I have read and followed the Guidance for submissions involving paid services.
• My contribution adds a new instruction, prompt, agent, skill, or workflow file in the correct directory.
• The file follows the required naming convention.
• The content is clearly structured and follows the example format.
• I have tested my instructions, prompt, agent, skill, or workflow with GitHub Copilot.
• I have run npm start and verified that README.md is up to date.

---

Description

The eval-driven-dev skill guides copilot through the full QA loop for LLM applications:

• Understand the code — read the codebase, trace the data flow, learn what the code is supposed to do
• Add instrumention — trace eval-related information to a local SQLite database
• Build a dataset — save representative traces as test cases
• Write and run eva-basedl tests — generate eval-based test files and get eval test results
• Investigate failures — look up the stored trace for test failure, diagnose, fix, repeat

The skill triggers when user ask for QA/Eval/Testing of LLM-based python code, or making changes that involves LLM calls.

Type of Contribution

• New instruction file.
• New prompt file.
• New agent file.
• New plugin.
• New skill file.
• New agentic workflow.
• Update to existing instruction, prompt, agent, plugin, skill, or workflow.
• Other (please specify):

---

Additional Notes

The skill installs and uses pixie-qa python package for trace, dataset and eval-test utilities. The package is open-sourced (MIT) here: https://github.com/yiouli/pixie-qa

Any data are stored locally and LLM calls during eval is also made directly from local machine.

---

By submitting this pull request, I confirm that my contribution abides by the Code of Conduct and will be licensed under the MIT License.

[Copilot]

Pull request overview

This PR introduces a new skill, eval-driven-dev, which guides Copilot through an eval-driven QA loop for Python LLM applications using the pixie-qa tooling, and wires it into the repo’s skills index.

Changes:

• Added the skills/eval-driven-dev skill with end-to-end instructions for instrumentation, dataset creation, and eval test execution.
• Added a bundled reference doc (references/pixie-api.md) describing pixie CLI and Python APIs.
• Registered the new skill in docs/README.skills.md.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File	Description
skills/eval-driven-dev/SKILL.md	Adds the core skill instructions and templates for running eval-driven QA workflows with pixie.
skills/eval-driven-dev/references/pixie-api.md	Provides a concise pixie API/CLI reference to support the skill’s workflow.
docs/README.skills.md	Adds the skill to the skills catalog so it shows up in generated listings.