dev-lead

Run an adversarial (red-team) review through the Antigravity CLI (`agy`, Gemini family). Use when the user asks for an adversarial review from Gemini/Antigravity, wants a second model family's opinion, or when another family's quota is tight and a different reviewer is needed. Pick a reviewer whose blind spots differ from the author's.

Delegate an implementation task to a headless Antigravity (`agy`) session in write mode, inside an isolated git worktree, then verify the result with a cross-family review pass before it merges. Use when the user wants a second model family's implementation attempt, wants to offload a well-scoped coding task, or explicitly asks to let agy implement something.

Run an adversarial (red-team) review through headless Claude Code (`claude -p`, plan mode) as the review leg. Use when a non-Claude family implemented the change and Claude is the chosen cross-family reviewer, or when a foreign lead (codex/agy) needs Claude as its review worker.

Delegate a well-scoped implementation to headless Claude Code (`claude -p`, acceptEdits) in an isolated git worktree, then verify independently and review cross-family before merge. Use when a foreign lead (codex/agy) needs Claude as its implementation worker, or when a Claude lead wants a separate Claude process with its own cwd and context.

Run a read-only Codex challenge review of a git diff through the Codex companion plugin, then verify each finding against the code. Use when the user asks for an adversarial or challenge review from Codex/GPT, wants a second opinion on a design or implementation, or needs the review gate after codex-implement.

Delegate a well-scoped implementation to OpenAI Codex through its Claude Code companion plugin in an isolated git worktree, then independently verify it with tests and adversarial review before a user-approved merge. Use when the user wants Codex to implement a bounded change, wants a second model's implementation attempt, or wants a safe Codex implementation-and-review workflow.

View and change the dev-lead leg roster — which model and effort each adapter runs, per role and per round. Use when the user wants to see or edit that roster.

Run a read-only challenge review through Cursor's CLI (cursor-agent) in ask mode. Use when the user asks for a review via Cursor, or when the cross-family rule needs a reviewer family the standalone CLIs cannot field — one paid adapter serving GPT, Claude, Grok, Kimi, Composer, and auto.

Dispatch an implementation task to Cursor's CLI (cursor-agent) in an isolated worktree, with the model pinned to whichever family the round's cross-family accounting needs. Use when the user asks Cursor to implement something, or to keep another family's quota free for review.

Orchestrate a feature or PR end-to-end as the lead — dispatch implementation to a delegate family based on task risk, model capability, and remaining quota; run bounded adversarial-review rounds with cross-family reviewers; verify everything independently; and hold the merge gate. Use when the user wants a whole feature/fix delegated and supervised, rather than a single implement or review pass.

Run a read-only challenge review through xAI's Grok Build CLI (Grok family). Use when the user asks for an adversarial or second-opinion review from Grok, or when the cross-family rule needs a reviewer family that GPT/Gemini/Claude/DeepSeek legs cannot provide — a paid-pool leg, tier peer of codex and agy.

Dispatch an implementation task to xAI's Grok Build CLI (Grok family) in an isolated worktree. Use when the user asks Grok/Grok Build to implement something, or when the paid pools' other families are reserved for review and a Grok write leg keeps the cross-family accounting clean.

Run an adversarial (red-team) review through the OpenCode CLI's free model pool (muse-spark, Nemotron, stealth models) under a machine-enforced read-only permission config. Use when a HIGH-risk change needs a second cross-family reviewer without spending paid quota, or as the cheap extra pair of eyes beside a primary review.

Delegate a well-scoped implementation to the OpenCode CLI's free model pool in an isolated git worktree with a machine-enforced no-push permission config, then verify independently before merge. Use for LOW-risk mechanical work where the free pool is the cheapest capable tier, when paid quota should be preserved, or when the user asks to let opencode implement something.