Skip to content

feat(pstack): use Opus 5.5 for design and review roles - #22

Merged
BramVR merged 1 commit into
mainfrom
codex/opus-review-roles
Sep 23, 2026
Merged

BramVR merged 1 commit into
mainfrom
codex/opus-review-roles

Conversation

@BramVR

@BramVR BramVR commented Sep 23, 2026

Copy link
Copy Markdown
Owner

Use Claude Opus 5.5 high instead of Astra low for Architect, Opus xhigh instead of Fable for Interrogate, and add Opus high to Arena's judge pool. Route Claude judges/reviewers through claude -p.

Optional Claude Autoreview now defaults to Opus 5.5 high, with a Claude Code 2.1.280 minimum version for primary and fallback selections. Astra remains the primary Codex reviewer and the single hardest-task model.

Validation: 106 skills validated; defaults self-test, four version tests, mirror tests, shell syntax and downloader smoke passed. Autoreview suite: 254 tests, one skip; two Java-dependent tests excluded because no Java runtime is installed. Previous live Opus CLI probe passed. Manual diff review completed.

Copilot AI lite review requested due to automatic review settings September 23, 2026 21:52
@BramVR
BramVR merged commit 33bdfe1 into main Sep 23, 2026
2 checks passed

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Copilot review overview

🟢 Approval recommended

The changes are consistent across role configs/docs, and the new Opus 5.5 minimum-version enforcement is implemented with targeted test coverage.

Review effort: Lite
Findings: None

What changed in this PR

Updates the pstack skill-model routing to prefer Claude Opus 5.5 for key design/review roles, while keeping Codex Astra as the primary Codex reviewer, and hardens autoreview’s Claude CLI version gating to enforce the Opus 5.5 minimum.

Changes:

  • Switch default/interrogate/architect/arena cross-judge model selections to include claude-opus-5-5 at the intended effort levels.
  • Route Claude reviewers/judges via claude -p ... --effort ... in skill docs where applicable.
  • Enforce and test Claude Code >= 2.1.280 when Opus 5.5 is selected (primary or fallback) in autoreview.
File Description
skills/​setup-pstack/​SKILL.md Updates the setup template’s role defaults to include Opus 5.5 for cross-judge, architect, and interrogate roles.
skills/​poteto-mode/​references/​models.md Aligns the canonical role→model configuration to include Opus 5.5 in the same roles/panels.
skills/​interrogate/​SKILL.md Switches the default Claude reviewer to Opus 5.5 and documents claude -p routing + minimum version.
skills/​autoreview/​tests/​test_autoreview_hardening.py Adds coverage for Opus 5.5 requiring Claude Code >= 2.1.280 for primary/fallback selection.
skills/​autoreview/​SKILL.md Updates autoreview documentation defaults and examples to Opus 5.5 at high effort.
skills/​autoreview/​scripts/​autoreview Changes Claude defaults to Opus 5.5/high and adds a stricter Opus 5.5 CLI version floor check.
skills/​arena/​SKILL.md Expands default cross-judge pool to include Opus 5.5 and clarifies transport parity with Phase B.
skills/​architect/​SKILL.md Replaces the default Astra-low architect runner with Opus 5.5 high, keeping the 3-candidate constraint.
CHANGELOG.md Records the Opus 5.5 role upgrades and the new minimum-version check.

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants