From 268546cc6e9600971354c046e7bb4d1669f1c6fe Mon Sep 17 00:00:00 2001 From: Claude Date: Fri, 24 Jul 2026 20:02:00 +0000 Subject: [PATCH 1/4] Migrate Claude model IDs to the 5 generation Move the frontend reviewer subagent to Opus 5 (deep bug-finding is the highest-value step in the review) and the CI review orchestrator to Sonnet 5 (diff classification, subagent dispatch, comment posting). Both replace claude-sonnet-4-6. No other API surface changes are needed: neither call site sets budget_tokens, sampling parameters, or assistant prefills, and effort: high remains valid on Opus 5. Co-Authored-By: Claude Opus 5 Claude-Session: https://claude.ai/code/session_01NkKxkLYPzNDTWUak4Wsb8D --- .claude/agents/uniplan-web-fe-reviewer.md | 2 +- .github/workflows/claude-code-review.yml | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/.claude/agents/uniplan-web-fe-reviewer.md b/.claude/agents/uniplan-web-fe-reviewer.md index 17b20f8..f394069 100644 --- a/.claude/agents/uniplan-web-fe-reviewer.md +++ b/.claude/agents/uniplan-web-fe-reviewer.md @@ -1,7 +1,7 @@ --- name: uniplan-web-fe-reviewer description: uniplan-web frontend reviewer. Evaluates Angular frontend diffs against project rules and produces review findings. Read-only — cannot edit code. System prompt carries the distilled reviewer context. Use for PR review flows and security passes on frontend code. -model: claude-sonnet-4-6 +model: claude-opus-5 effort: high tools: Read, Glob, Grep, Bash, Skill, ToolSearch --- diff --git a/.github/workflows/claude-code-review.yml b/.github/workflows/claude-code-review.yml index a94f910..54476d0 100644 --- a/.github/workflows/claude-code-review.yml +++ b/.github/workflows/claude-code-review.yml @@ -105,7 +105,7 @@ jobs: # The prompt posts via `gh pr review`, diffs via `git`, dispatches the # uniplan-web-fe-reviewer subagent (Task) and runs /review-pr (Skill). claude_args: >- - --model claude-sonnet-4-6 + --model claude-sonnet-5 --allowedTools "Bash(gh pr review:*),Bash(gh pr diff:*),Bash(gh pr view:*),Bash(git:*),Task,Read,Glob,Grep,Skill" prompt: | Run the project's `/review-pr` skill against PR #${{ inputs.pr_number }} From 56b924759623227d71b722090a912c6e4a21ebe3 Mon Sep 17 00:00:00 2001 From: Claude Date: Fri, 24 Jul 2026 20:16:27 +0000 Subject: [PATCH 2/4] Tune review flow for the Opus 5 / Sonnet 5 reviewers Follow-up to the model migration, covering prompt-side adjustments the model IDs alone did not address. - Reviewer: cap finding length at one to two sentences. Opus 5 writes longer user-facing text than Sonnet 4.6, and this output is posted verbatim as a PR comment. Verbosity responds to prompting, not effort. - Reviewer: effort high -> medium. Opus 5 holds review accuracy at lower effort; revert to high if recall drops on real PRs. The skill/workflow CI override and the review-only guardrail this repo already had are what the uniplan side was brought in line with. Co-Authored-By: Claude Opus 5 Claude-Session: https://claude.ai/code/session_01NkKxkLYPzNDTWUak4Wsb8D --- .claude/agents/uniplan-web-fe-reviewer.md | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/.claude/agents/uniplan-web-fe-reviewer.md b/.claude/agents/uniplan-web-fe-reviewer.md index f394069..4e47de7 100644 --- a/.claude/agents/uniplan-web-fe-reviewer.md +++ b/.claude/agents/uniplan-web-fe-reviewer.md @@ -2,7 +2,7 @@ name: uniplan-web-fe-reviewer description: uniplan-web frontend reviewer. Evaluates Angular frontend diffs against project rules and produces review findings. Read-only — cannot edit code. System prompt carries the distilled reviewer context. Use for PR review flows and security passes on frontend code. model: claude-opus-5 -effort: high +effort: medium tools: Read, Glob, Grep, Bash, Skill, ToolSearch --- @@ -113,6 +113,8 @@ Tests: If no issues found, state: **"No issues found."** +**Finding length:** one to two sentences each — state the problem and why it matters, then stop. Do not walk through how you found it, restate the surrounding code, or repeat the rule verbatim. Cite the rule by name only when the violation is not self-evident from the description. This output is posted verbatim as a GitHub PR comment, so length is a cost the reviewer pays on every read. + ## Workflow specifics - **PR review** — invoking prompt provides worktree path + base/head. Context line: `PR # review`. From df590320454d68cdf28a1d5c48da5f14bbcecaba Mon Sep 17 00:00:00 2001 From: Claude Date: Fri, 24 Jul 2026 20:21:19 +0000 Subject: [PATCH 3/4] Restore effort: high on the frontend reviewer Reverts the medium step-down from the previous commit. Review recall is worth more here than the token saving; the finding length cap is unaffected. Co-Authored-By: Claude Opus 5 Claude-Session: https://claude.ai/code/session_01NkKxkLYPzNDTWUak4Wsb8D --- .claude/agents/uniplan-web-fe-reviewer.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/.claude/agents/uniplan-web-fe-reviewer.md b/.claude/agents/uniplan-web-fe-reviewer.md index 4e47de7..677741d 100644 --- a/.claude/agents/uniplan-web-fe-reviewer.md +++ b/.claude/agents/uniplan-web-fe-reviewer.md @@ -2,7 +2,7 @@ name: uniplan-web-fe-reviewer description: uniplan-web frontend reviewer. Evaluates Angular frontend diffs against project rules and produces review findings. Read-only — cannot edit code. System prompt carries the distilled reviewer context. Use for PR review flows and security passes on frontend code. model: claude-opus-5 -effort: medium +effort: high tools: Read, Glob, Grep, Bash, Skill, ToolSearch --- From 245f4d3b2d8cc5c6805056e4c67bab828fc657c7 Mon Sep 17 00:00:00 2001 From: Stoyan Atanasov <47278663+s100ian@users.noreply.github.com> Date: Thu, 30 Jul 2026 16:03:26 +0300 Subject: [PATCH 4/4] Update model from claude-sonnet-5 to claude-opus-5 --- .github/workflows/claude-code-review.yml | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/.github/workflows/claude-code-review.yml b/.github/workflows/claude-code-review.yml index 54476d0..c684792 100644 --- a/.github/workflows/claude-code-review.yml +++ b/.github/workflows/claude-code-review.yml @@ -105,7 +105,7 @@ jobs: # The prompt posts via `gh pr review`, diffs via `git`, dispatches the # uniplan-web-fe-reviewer subagent (Task) and runs /review-pr (Skill). claude_args: >- - --model claude-sonnet-5 + --model claude-opus-5 --allowedTools "Bash(gh pr review:*),Bash(gh pr diff:*),Bash(gh pr view:*),Bash(git:*),Task,Read,Glob,Grep,Skill" prompt: | Run the project's `/review-pr` skill against PR #${{ inputs.pr_number }}