AI Engineering

OpenCode Delegate hands a bounded coding task to the OpenCode CLI as a background implementer, then lets you review its diff and commit it yourself. Use it when the user wants to delegate implementation work to OpenCode or run a queue of coding tasks while staying the reviewer.

Pi Delegate delegates a bounded coding task to the Pi coding agent CLI as a background implementer, then lets the orchestrator review the diff and commit. Use when the user wants to hand implementation work to Pi or run a queue of coding tasks while staying the reviewer.

Qoder Delegate dispatches bounded coding tasks to the Qoder CLI as a background implementer, then lets you review the diff and land the changes yourself. Use it when you want Qoder to implement, fix, refactor, or run a queue of coding tasks while you remain the reviewer.

Vibe Delegate delegates a bounded coding task to the Mistral Vibe CLI and then reviews and lands the diff. Use it when you want to hand implementation work to Vibe instead of writing code inline.

Warp Delegate delegates coding tasks to the Warp Agent CLI (oz) as a background implementer, then reviews its diff and lands it. Use it when you want to hand implementation work to Warp or run a queue of tasks through Warp while staying the reviewer.

Improve Harness guides one explicit, evidence-backed improvement to a repository's agent guidance, tools, runbooks, or validation after observed reusable agent friction. Use only when the user invokes $improve-harness or explicitly asks to improve the Harness.

Onboard Repository inspects an unfamiliar or brownfield repository, traces one real operational path, and proposes evidence-backed improvements that help future agents work independently. Use it when asked to onboard, map, assess, or backfill agent-facing repository guidance.

Claude Code Authoring Kit is a comprehensive reference for creating and configuring Claude Code extensions, covering Skills, sub-agents, plugins, slash commands, hooks, memory, settings, sandboxing, headless mode, and advanced agent patterns.

Cross-Model Audit Convergence documents how plan-auditor and sync-auditor agents invoke the audit_multi MCP tool to run code review across codex and GLM backends in parallel, converge their verdicts with Claude, and fold per-backend verdicts and a disagreement flag into audit output. Use it when a project sets audit_model to multi.

MoAI Foundation Core provides foundational principles for MoAI-ADK, including the TRUST 5 quality framework, SPEC-First DDD methodology, delegation patterns, progressive disclosure, and token budget management. Use for TRUST 5 gates, SPEC workflow, or context optimization.

MoAI Foundation Thinking provides structured frameworks for critical evaluation, diverge-converge brainstorming, and deep questioning to analyze proposals, generate solutions, and explore complex problems. Use it for ideation, strategic analysis, architecture decisions, and deep reasoning.

MoAI Harness Learner coordinates the harness learning subsystem, producing Tier 4 auto-update proposal payloads for the MoAI orchestrator to surface via AskUserQuestion, and manages Apply/Rollback workflows.