From 3bd6283112c293b0d6a25f6f1c30b4d68b903b1a Mon Sep 17 00:00:00 2001 From: Nathan Heskew Date: Mon, 20 Apr 2026 10:39:58 -0700 Subject: [PATCH] =?UTF-8?q?workflow:=20bump=20max-turns=2016=20=E2=86=92?= =?UTF-8?q?=2024=20for=20headroom=20on=20substantive=20PRs?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit PR #43 (Harper v5 migration) hit 16 turns with zero denials and $0.87 spent, posting one valid finding before cutoff. Bumping to 24 gives room to complete the review without sliding into Opus-cost territory. Sonnet at 24 turns is still bounded (~$1.30 worst case, historical runs were ~$0.05/turn) and concurrency-1 + 10min job timeout stay unchanged. For PRs that still cap out at 24, the right move is to split the PR, not keep raising turns. Co-Authored-By: Claude Opus 4.7 (1M context) --- .github/workflows/claude-review.yml | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/.github/workflows/claude-review.yml b/.github/workflows/claude-review.yml index da5b529..1ab7b75 100644 --- a/.github/workflows/claude-review.yml +++ b/.github/workflows/claude-review.yml @@ -37,7 +37,7 @@ jobs: show_full_output: true # TEMP: keep on during calibration so tool denials are visible; revert once reviews run cleanly claude_args: | --model claude-sonnet-4-6 - --max-turns 16 + --max-turns 24 --allowedTools "mcp__github_inline_comment__create_inline_comment,Bash(gh pr comment:*),Bash(gh pr diff:*),Bash(gh pr view:*),Read,Grep,Glob" prompt: | REPO: ${{ github.repository }}