Add a social approval skill evaluation

[kriptoburak]

Summary

• Add an evaluation scenario for approval-gated social publishing through an MCP workflow.
• Update the README evaluation count and tree so the new fixture is discoverable.
• Assert dry-run default behavior, exact approval before publishing or scheduling, and no secret leakage.

Validation

• node -e "JSON.parse(require('fs').readFileSync('evaluations/eval-social-approval-skill.json','utf8')); console.log('json ok')"
• npx -y markdown-link-check README.md --quiet found one pre-existing transient homepage failure: https://lijinnair.com returned 503. Re-run with only that known URL ignored passed.
• git diff --check

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 2f343f5905

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]

Assert side-effect skills are not auto-invoked

For this approval-gated publishing scenario, the repo's own Step 3 requires disable-model-invocation: true for workflows with controlled side effects such as sending/publishing. This fixture only checks that front matter is standard and that the prose asks for approval, so a generated social-posting skill can pass the evaluation while remaining auto-discoverable/invocable, undermining the safety condition the new eval is meant to cover. Add an expected behavior for the correct invocation flag here.

Useful? React with 👍 / 👎.

[kriptoburak]

Pushed 7ec4701 to add disable-model-invocation: true to the social publishing eval fixture. Validation: python3 -m json.tool evaluations/eval-social-approval-skill.json, git diff --check.