Bootstrap a Claude Code agent harness in the current project — CLAUDE.md, AGENTS.md, .claude/agents/, hooks, plugin wiring, all in one invocation. Use when starting a new project or adopting Claude Code on an existing one without an existing CLAUDE.md. (Renamed from /harness-init in v0.2.0)
Cost-unrestricted multi-agent pipeline. Drives intent→plan→wave-dispatch→gate→PR over the .claude/agents/ roster, with optional --loop until a shell break-condition succeeds (bounded by --max-iter and --max-cost). Requires /agent-init scaffolding.
Scaffold AGENTS.md, .codex/skills/, an operational task ledger, a repo-local Codex policy hook, and an operational Codex config snippet for a new or existing project. Use --lite to opt out of the heavy operational artifacts.
Codex CLI port of /agent-all (intent → plan → wave-dispatch → gate → PR). Current Codex hooks do not expose the older agent-dispatch surface, so Phase 3 uses sequential `.codex/skills/<role>` invocations. The local phase files in this skill are the runnable Codex workflow contract.
Codex CLI port of /thrift — cost-conscious long-session optimisation. TOML hooks in ~/.codex/config.toml, OpenAI cost table, session-priming variant of Phase 4. Same six-phase pipeline as harness-thrift (CC); see plugins/harness-thrift/skills/thrift/SKILL.md for source-of-truth.
Codex CLI port of /debug — six-phase debugging workflow with hypothesis state persistence, structured error parsing, and git/input bisection. Use run /debug to start an investigation from a failing command; run /debug --resume to continue across sessions. Writes a durable log to docs/debug/<date>-<slug>.md at end. Wraps (does not replace) superpowers:systematic-debugging when that skill is installed.
Six-phase debugging workflow with hypothesis state persistence, structured error parsing, and git/input bisection. Use /debug to start an investigation from a failing command; /debug --resume to continue across sessions. Writes a durable log to docs/debug/<date>-<slug>.md at end. Wraps (does not replace) superpowers:systematic-debugging when that skill is installed.
Codex CLI port of /visual-qa — Playwright MCP capture matrix + per-image LLM analysis + diff vs prior run. Supports `declared` and `comprehensive` modes (crawl + DOM walk auto-discovery, shallow click, baseline-relative verdict). Phase 3 uses sequential `.codex/skills/visual-qa-page` dispatch because current Codex hooks do not expose the older agent-dispatch surface. The local phase files in this skill are the runnable Codex workflow contract.