skills readme was getting junk files

[aaronpowell]

Pull Request Checklist

• I have read and followed the CONTRIBUTING.md guidelines.
• 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

---

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

---

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 aims to prevent the generated Skills README from listing “junk”/overly-verbose bundled asset paths by changing how skill asset paths are discovered and then updating the generated docs/README.skills.md output accordingly.

Changes:

• Restricts skill asset recursion in parseSkillMetadata() to specific top-level directories.
• Updates the generated skills table to show some bundled assets as directory entries (e.g., templates, examples, scripts/cache-generator) instead of enumerating every file.

Reviewed changes

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

File	Description
eng/yaml-parser.mjs	Changes skill asset enumeration logic (affects README generation and other consumers of assets).
docs/README.skills.md	Regenerated skills README reflecting the new bundled-asset listing behavior.

Comments suppressed due to low confidence (1)

eng/yaml-parser.mjs:146

• The inline comment says this lists "all files" and "recursing through subdirectories", but the updated logic only recurses into a small allowlist of directory names and may add directories themselves to the assets list. Please update the comment to match the new behavior so future readers don’t assume it’s a full recursive file walk.

      // List bundled assets (all files except SKILL.md), recursing through subdirectories
      const getAllFiles = (dirPath, arrayOfFiles = []) => {

---

You can also share your feedback on Copilot code review. Take the survey.

[Copilot]

Changing getAllFiles to only recurse into references/assets/scripts means metadata.assets can now include directory entries (e.g., templates, examples, scripts/cache-generator) instead of listing all files. Downstream consumers treat assets as files (e.g., validate-skills checks each asset path size via statSync), so directories will bypass per-file size validation and also skew asset counts in website metadata. Consider keeping parseSkillMetadata().assets as a full file list (and collapse/group only when rendering the README), or update the validation/consumers to recurse into directories when an asset entry is a folder.

This issue also appears on line 145 of the same file.

See below for a potential fix:

        files.forEach((file) => {
          const filePath = path.join(dirPath, file);
          const stat = fs.statSync(filePath);
          if (stat.isDirectory()) {

[Copilot]

Quote style is inconsistent with the rest of this file (which primarily uses double quotes). To keep formatting consistent and avoid lint/prettier churn, consider switching ['references', 'assets', 'scripts'] to use double-quoted strings.

Suggested change

	const assetPaths = ['references', 'assets', 'scripts'];
	const assetPaths = ["references", "assets", "scripts"];