sd0xdev/sd0x-dev-flow

Adr writes an architecture decision record for a feature in Context, Decision, Status, Consequences, and Alternatives. Use it to document why an architectural choice was made or to mark a prior decision as superseded.

Architecture Design produces 3-architecture.md from an existing tech spec or code context. It documents components, data flow, integration points, and architecture decisions for a feature.

Ask provides context-aware Q&A with automatic context gathering for codebase, git history, rules, docs, and other skills during development. Use it for quick questions when you need a sourced answer, not for code changes or deep research.

Best Practices Audit checks a codebase against industry standards and best practices, then produces a verdict, gap roadmap, and debate evidence. Use it when you need a conformance audit rather than broad research or general code review.

Bug Fix fixes bugs, issues, and regressions with a root-cause workflow. Use it to investigate, patch minimally, add a regression test, and run review gates before precommit.

Bump Version updates package.json, .claude-plugin/plugin.json, and the install-state manifest to the same version. Use it when a package or plugin version needs to be incremented in sync.

Claude Health Check audits .claude structure, naming, hooks, and plugin version drift, then reports issues with fix recommendations. Use it when verifying config hygiene or syncing installed assets.

Code Explore traces execution paths and explains how a feature or bug works from the codebase. Use it for quick architecture understanding, root-cause diagnosis, and single-perspective investigation reports.

Code Investigate performs deep code analysis with independent Claude and Codex passes, then merges the findings. Use it to trace how a feature works, map data flow, or confirm complex behavior.

Codex Architect provides architecture advice, design recommendations, and option comparisons for a project. Use it when designing features, evaluating an existing design, or getting a second opinion before implementation.

Codex Brainstorm runs independent Claude and Codex research, then an adversarial debate to compare solution options and feasibility. Use it for brainstorming, exhaustive exploration, and final consensus with action items.

Codex CLI Review runs Codex CLI on uncommitted or branch changes with full disk access. Use it for deep codebase review and severity-grouped findings, not for quick diff checks or documentation review.

Codex Code Review runs Codex-based review on pull requests and code changes, then returns severity-grouped findings and a merge gate. Use it for PR review, code audits, or a second opinion on changes.

Codex Explain produces structured explanations of code to help with understanding complex logic, tracing data flow, and onboarding to unfamiliar code. Use it when you need a walkthrough at brief, normal, or deep depth.

Codex Implement writes new features from specs using Codex exec and a review loop. Use it for feature implementation and new code, not for code review, bug fixes, or architecture advice.

Codex Review routes to a full code review workflow using Codex exec, including lint:fix and build checks. Use it for a second opinion on a diff with context preserved across the review loop.

Codex Review Branch runs a thorough automated review of all commits on a feature branch using Codex exec. Use it when you want branch-wide code review without lint or build checks.

Codex Review Doc routes to the document review workflow for reviewing docs with Codex exec. Use it when you need review loops with preserved context and ratings.

Codex Review Fast routes to a quick diff-only second opinion using Codex exec with no tests. Use it for lightweight review loops when you want comments without running lint or build checks.

Codex Security routes to the full OWASP Top 10 security review workflow with context preservation. Use it for security audits and review loops, not general code or dependency checks.

Codex Setup initializes sd0x-dev-flow for Codex CLI and other non-Claude agents by generating AGENTS.md, installing hooks, and copying runner scripts. Use it for first-time project setup or after updating skills.

Codex Test Gen generates new unit tests for specified functions using the Codex exec workflow. Use it when you want test cases created, not when reviewing existing coverage or writing integration tests.

Codex Test Review evaluates existing test coverage for sufficiency and suggests missing edge cases. Use it when reviewing tests rather than generating new ones.

Contract Decode decodes EVM revert data, function selectors, custom errors, and calldata from hex input. Use it when you have 0x-prefixed calldata, a selector, or an execution reverted message and need the signature or arguments.

Test Coverage Analysis assesses unit, integration, and E2E coverage for a feature and identifies missing cases. Use it to map source files to tests, find gaps, and recommend new tests.