Conversation
Replace the three public skills with one intent-matched entry skill that loads the requested mode and shared decision method. BREAKING CHANGE: Use work-through-decisions instead of the former skill names.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Why
Issue #228 tracks the GPT-6 Luna/medium eval sweep and follow-up on failing plugins. The first full pass found 86 failures across 361 direct-runner cases, so the results need a durable record and focused repairs. Refs #228.
What changed
Switch the default Codex eval route from Terra/medium to GPT-6 Luna/medium and record the full sweep with case-level assessments. Add semantic grading for saved artifacts, unify the discovery workflows, split skill creation from read-only auditing, allow explicitly delegated PR template selection, and clarify Langfuse attribution guidance and its negative eval boundary. Update the affected specifications, evals, and plugin versions.
Verification
The initial 361-case Luna/medium sweep completed; 275 passed and 86 failed, as recorded in docs/research/gpt-6-luna-evals-issue-228.md. Focused follow-ups in that report passed 5/5 for decision provenance, seven discovery boundaries, all 13 active skill-authoring cases in separate five-trial runs, and both Langfuse failures; the 12 create-pr cases passed a fresh single-trial regression. The latest Langfuse change also passed bun run check:docs, a Prettier check, and git diff --check. These focused samples do not establish a clean full-suite rerun.
Review notes
This is a draft because issue #228 still has unresolved plugin failures, including readiness-gate, and the original 86-failure overview remains historical. Review the runner's saved-artifact grader and the discovery and skill-authoring routing changes closely. Repository-wide pre-PR gates and cross-host behavior have not been established for the whole branch.
Checklist
CONTRIBUTING.md, including the contributionlicensing terms.
this change does not affect a capability invariant.
behavior.
change does not affect plugin content.
bun run check:python, or this change does not affect registeredPython packages or their repository quality infrastructure.