Repository navigation
feat: planner no longer capped at 25 turns, add vanilla comparison runner - #196
Merged
Merged
Conversation
…ures, validates numeric flags
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What this does
Large planning jobs no longer get cut off after 25 turns. It also adds a script that scores the eval cases with the full cc-settings setup against plain Claude Code, so skills and rules that no longer beat the bare model show up as deletion candidates.
Summary
agents/planner.md: dropmaxTurns: 25(effort staysxhigh).src/scripts/eval-ablate.ts: builds a throwaway plugin (skills, agents, src, standards injected through SessionStart hooks) and runsclaude plugin eval --ablation with-without. Cases with no comparison are listed separately and not counted as "vanilla matches". Numeric flags are validated..claude/AGENTS.mddoc line.eval-ablate.ts, then the rest.Test Plan
bun run typecheckcleanbun test: 2129 pass, 0 failbun src/scripts/eval-ablate.ts --runs abcexits 1 with a clear messageci-gategreenbun src/scripts/eval-ablate.ts --runs 3 --max-cost-usd 35(started after the PR opens)