Lifecycle phasesv0.1.0

auto-review

Fans a change across independent review lenses and reports only findings that survive adversarial refutation.

WORKFLOW / 05/auto-review [scope — a diff/branch/PR, or path] (default: the current branch diff)
  1. 01

    Pin the scope, baseline, diff, and review budget.

  2. 02

    Fan out correctness, security, performance, maintainability, testing, and contract lenses.

  3. 03

    Deduplicate raw findings without erasing distinct failure scenarios.

  4. 04

    Put every finding through an independent majority-refute panel.

  5. 05

    Report survivors, open quorum gaps, and rejections; fix only when requested.

operating envelope

Use it deliberately.

The shortest useful definition of where this skill helps—and where it should stay out of the way.

Use when

  • A branch, pull request, diff, or module needs broad review before merge.
  • False positives need to be filtered before they consume engineering attention.
  • Confirmed blockers should optionally enter a bounded, scoped fix loop.

Do not use when

  • You need the feature implemented rather than reviewed.
  • You want behavior-neutral cleanup; use auto-simplify.
  • You have not explicitly asked for fixes and expect a read-only review.

output contract

What must be true when it stops.

Produces

  • Verified findings grouped by Blocker, Concern, Nit, and FYI
  • Coverage report for every review dimension
  • Unverified quorum-gap ledger
  • Refuted-finding rejection ledger
  • Optional bounded fixes for confirmed issues

Guarantees

  • Only adversarially verified findings count or drive edits.
  • A review dimension that did not run is an explicit coverage gap.
  • Severity is calibrated to the actual failure scenario.
  • Fixes are opt-in, minimal, scoped, and re-verified.
  • A clean verdict is impossible while a dimension or verification quorum is missing.

Honest failure states

  • coverage_gap — one or more review dimensions did not return evidence
  • unverified — a finding panel died below quorum
  • blocked — verified blocker findings remain open
  • fix_exhausted — an opted-in repair loop did not converge

enforcement

Know which claims are executable.

Scripts are linked when this skill owns deterministic machinery. The remaining rules are explicit operating contracts and must not be marketed as hooks.

install and invoke

One skill, three honest surfaces.

skills.sh

Available

Universal skill installation, including Claude Code and Codex.

npx skills add https://github.com/ulpi-io/skills-autonomous-engineering --skill auto-review

Claude Code plugin

Available

Install the full plugin, then invoke /auto-review or let routing select it from context.

Claude installation →

Codex plugin

Work in progress

A work in progress on the codex-native-plugin branch — build it, start a new session, then invoke $autonomous-engineering:auto-review.

Codex installation →