feat: add Claude effort dispatch and refresh model guidance - #315
Merged
Merged
Conversation
tkstang
changed the title
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Claude managed subagents now select model and reasoning effort independently. OAT resolves and materializes exact reviewer and phase-implementer variants, verifies generated frontmatter against launch intent, and preserves model-only and inherited paths. Planning prompts ask about gates for the active workflow and its downstream implementation without prompting for unrelated planning modes.
The model catalog and recommended dispatch ladders are refreshed. Claude Opus 5.5 replaces active Opus 4.8/5.0 routes. Claude Balanced now has one candidate, Opus 5.5 low instead of Sonnet 5 high; High also admits Opus 5.5 low for bounded intelligent recon, alongside medium and high for more substantive work. Codex intelligent recon stays at Sol medium. GPT-6 Sol/Luna are added to supported Codex targets and preferred recommendations while GPT-5.6 remains available. Existing explicit dispatch cells are preserved. Exact Codex Sol
ultracandidates resolve without expanding the project-wide scalar ceiling beyondmax, and a lower--preferredeffort remains honored under anultrafrontier candidate.Cursor Opus 5.5 has five approved mappings backed by redacted native desktop probe records stored with the Cursor codec fixtures. GPT-6 remains outside Cursor's bundled preferred ladder pending native availability and pin evidence. The prior
claude-sonnet-5-highremains a non-catalogue compatibility alias for explicit cells, covered by a resolver-level regression test;claude-sonnet-5-thinking-highis the canonical catalogue ID.The Updating Model Guidance page identifies canonical catalogs, recommendation assets, provider references, generated views, model-ID and pin evidence, adoption semantics, and validation steps for future updates. Bundled skill and agent versions and the five public packages are bumped; the shipped dispatch recommendation is regenerated.
Validation on this branch: focused CLI suite 564/564, followed by
pnpm check,pnpm type-check,pnpm test(7,528 CLI tests executed),pnpm build,pnpm run check:skill-bumps,pnpm release:check-versionsagainst fetchedorigin/main,pnpm release:validate,pnpm build:docs,pnpm lint, andpnpm format, all passing in CI order. The latest configured independent gate reviewed2cfb86bf7and passed its High threshold with 0 Critical/High/Medium and 2 Low findings; both Low items were addressed in subsequent commits. The final Claude ladder change and those fixes were locally validated; at the user's request, no additional independent review was run after them.Project artifacts: discovery, plan, implementation, reviews.