Release v3.31.0

[atomantic]

Release v3.31.0

Released: 2026-08-16

Highlights

• Per-reviewer reasoning effort suffix (~effort=<level>) on --review-with: Any slot in a --review-with list across all review commands (/do:pr, /do:release, /do:review, /do:better, /do:better-swift, /do:depfree, /do:rpr, /do:config) can now specify its reasoning effort level (low, medium, high, xhigh, max) — e.g. --review-with=codex[gpt-5.6-luna]~effort=max~opt, claude~effort=high~max=2.
• Suffix chaining & saved defaults: Suffixes (~opt, ~max=<n>, ~effort=<level>) chain in any order and ride through /do:config saved defaults in canonical order (~opt -> ~max=<n> -> ~effort=<level>).
• Updated documentation & workflow examples: README.md includes a Typical Developer Loop example (plan-task -> next -> pr) and explicit assistant invocation syntax details for Claude Code (/do:*), Codex ($do-*), Antigravity (/do-*), OpenCode (/do-*), and Grok Build (/do-*).

Details

• Reviewer Loop Threading: Per-reviewer reasoning effort is forwarded as {REVIEW_EFFORT} to local agent review loops (local-agent-review-loop.md) and as {OLLAMA_EFFORT} to local Ollama review loops (ollama-review-loop.md).
• CLI Flags & Prompt Construction: Local agent invocations pass ${EFFORT_FLAG[@]+"${EFFORT_FLAG[@]}"} (--effort <level>) where supported, and Ollama review loops append reasoning effort instructions directly to the prompt.