Standalone
No MCP, tracker, coordinator, worker hierarchy, or shared prompt harness.
Claude Code + Codex · standalone by design
A compact engineering toolkit that reminds capable models about the hard parts: evidence, boundaries, independent viewpoints, reversible experiments, and residual risk.
No MCP, tracker, coordinator, worker hierarchy, or shared prompt harness.
Findings, verdicts, and changes are tied to observable evidence and limitations.
One canonical SKILL.md works in both Claude Code and Codex.
The catalog
The first digit identifies the plugin family. The second identifies a skill inside it.
Plugin 01
Independent, read-only gates for plans and delivered code.
Tests a proposed implementation against repository evidence, risks, and missing decisions.
Reviews completed code through independent, risk-selected perspectives and a strict delivery verdict.
Plugin 02
Read-only audits that separate confirmed evidence from uncertainty.
Trust, coverage, structure, freshness, and factual accuracy.
Security boundaries, delivery gates, maintainability, and diagnosability.
Portfolio value, risk coverage, isolation, and oracle strength.
Ownership, contracts, dependency topology, and structural fitness.
Queries, transactions, consistency, concurrency, and resource lifecycle.
Plugin 03
Bounded changes survive only when measurements justify keeping them.
Profile, experiment, compare, and keep only reproducible improvements.
Upgrade in reversible batches with compatibility and regression evidence.
Simplify or replace a bounded capability only when net value is proven.
Compare alternatives under a frozen experiment contract.
Plugin 04
Risk-shaped strategy and executable acceptance evidence.
Maps requirements and defect classes to the right test boundary and oracle.
Creates reproducible tests without silently repairing product defects.
Plugin 05
External viability before implementation enthusiasm.
Compares demand, acquisition, competition, economics, and validation cost using current evidence.
Plugin 06
Optional, repository-aware workflows for maintainers who review skills and publish code, releases, or project news.
Checks standalone contracts, triggers, behavior, and configured distribution surfaces before publication.
Validates, commits, pushes, and verifies approved changes from their authorized remote.
Builds an evidence-backed tagged release and waits for approval before publication.
Drafts, fact-checks, and publishes GitHub Discussions announcements after approval.
Plugin 07
Independent workflows that create durable architecture artifacts and communicate only through optional repository documents.
Establishes measurable architecture drivers and constraints in one project baseline.
Maps implemented architecture from repository evidence without issuing an audit verdict.
Turns requirements, estimates, boundaries, and tradeoffs into a target design.
Preserves one significant decision with alternatives, consequences, and review triggers.
Creates evidence-backed current or target views with explicit scope and notation.
Plans reversible coexistence, data movement, cutover, rollback, and old-path removal.
Install only what you need
/plugin marketplace add levnikolaevich/claude-code-skills
/plugin install review-suite@levnikolaevich-skills-marketplace
codex plugin marketplace add levnikolaevich/claude-code-skills
codex plugin add review-suite@levnikolaevich-skills-marketplace
Replace review-suite with codebase-audit-suite, optimization-suite, testing-suite, product-discovery-suite, maintainer-suite, or architecture-suite.