Add SkillOpt-ready metadata to decision traces

[ruicore]

Summary

• Extend decision-trace-writer so durable traces can include compact SkillOpt-ready extraction metadata.
• Add a machine-readable SkillOpt 数据捕获 block with candidate skill, validation, evaluator, fixture, privacy, and split-suggestion fields.
• Preserve the original worth-recording gate: do not create traces solely as raw training data.
• Add explicit privacy and split-currency rules so session-derived traces default to training candidates unless curated.

Rationale

This keeps decision traces as agent-facing project memory while making them easier to mine later into task items, eval cases, benchmark fixtures, or SkillOpt-style skill-optimization data.

Validation

Not run locally from this connector. The change is Markdown-only; recommended follow-up is:

python scripts/validate_skills.py

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 7d8b352d12

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Use a longer fence for the nested YAML template

This ```yaml fence sits inside the ```markdown example opened above, so Markdown parsers treat it as the closing fence for the whole trace template rather than as a nested YAML block. As a result, the SkillOpt fields render as ordinary Markdown and the following backticks become an empty fence, which makes the copyable trace template broken/confusing; use a longer outer fence such as ````markdown or indent/escape the inner YAML fence.

Useful? React with 👍 / 👎.