fix: let explain and pov support calling workflows

[tmchow]

Summary

ce-explain and ce-pov can contribute to planning, brainstorming, and PR writing without assuming a person is waiting for a teaching document or follow-up menu. Delivery follows the requested use: an answer, text for another document, or a standalone explainer.

ce-explain owns understanding how something works and investigating its historical rationale. ce-pov owns the recommendation and calls for an explanation only when unresolved understanding could change that recommendation. Planning and brainstorming use the same conditional handoff; no new skills or return-mode flags are introduced.

Standalone teaching retains offline HTML, portable Markdown, source attribution, prose that stands without diagrams, and static exercises with answers. Oracle still requests independent model opinions. Publication remains separate from completing an explanation and retains its consent requirements.

Validation

• 3,904 repository tests passed, plus release and plugin validation.
• 40 real CLI invocations exercised PR-body explanations, planning judgments, teaching artifacts, and native Oracle activation, including iterations after failures.
• Tested Claude Opus 5 High and Fable 5.1 Low, Codex 5.6 Sol Medium and Astra Low, Grok 4.6 Medium, and Cursor Composer-2.5.
• All six returned content to calling workflows without unnecessary human checkpoints or implementation. Opus and Astra selected ce-pov from “Oracle this” and completed real peer calls.
• Reruns corrected Composer's invalid Markdown fence labels and invented audience metadata.

Grounding is not uniformly passing. Astra consistently preserved unknowns in the inspected scenarios. Sol's latest outputs passed, but an earlier repeat overclaimed. Opus, Fable, Grok, and Composer still made some unsupported claims about uninspected queue behavior despite revisions. A correct recommendation was not counted as proof that its supporting claims were correct.

These were bounded fixture tests, not a complete lfg run or application-repository evaluation. HTML structure was inspected; browser rendering was not tested. The final removal of a duplicate evidence instruction was rerun on Fable and Composer, rather than repeating the entire matrix.

Security Disclosure

Explanation delivery no longer requires publication or a destination selection. Existing public-publishing confirmation and offline-artifact restrictions remain. No new shell execution, credentials, dependencies, or publisher integrations were introduced.

Agent Disclosure

• Model: Codex CLI · GPT-6

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-09T01:51:40.633712Z	7a7b9b2	PR opened

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" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 7a7b9b24da

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".