orchestration-methodology

Orchestrate delegated coding tasks through a six-phase Architect/Executor methodology.

Updated Jul 30, 2026
One-click install
npx skills add https://github.com/primax79/ai-architect-executor --skill orchestration-methodology-primax79
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: orchestration-methodology
Source: https://github.com/primax79/ai-architect-executor/tree/main/plugins/architect-side/skills/orchestration-methodology
Command: npx skills add https://github.com/primax79/ai-architect-executor --skill orchestration-methodology-primax79

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Delegating implementation work to a fast coding agent often fails silently: tasks race each other in the same directory, reports claim success without real verification, and merges land on the wrong branch. This Skill gives the orchestrating AI a disciplined, protocol-agnostic methodology for delegating work to an Executor agent without losing control of quality or state. ## Core Features & Use Cases - Six-phase delegation lifecycle: discovery-first exploration, worktree isolation, non-blocking parallel delegation, complexity-scaled monitoring, execution-based verification, and closure with defect telemetry. - Risk-calibrated monitoring and intervention: cross-check OS-level liveness against cost, tokens, and plan activity; cancel drifting tasks and resume corrective follow-ups in the same session. - Real verification and integration gates: never trust the Final Report alone - run builds, tests, and a full consuming-application build after every merge, and track task state as a path (delivered, verified, integrated) rather than a done/not-done label. - Use Case: You ask an AI to implement three independent features. The Skill guides the orchestrator to isolate each in its own git worktree, dispatch them in parallel with distinct ports, poll progress proactively, execute the results to verify, and merge with an explicit checkout of the target branch. ## Quick Start Ask the AI to orchestrate the implementation of a feature by delegating it to an Executor agent following the orchestration-methodology phases.

Frequently Asked Questions about orchestration-methodology

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I delegate coding tasks to an AI agent safely?▼

Isolate each task in its own git worktree before dispatching, since delegation has no locking by default and concurrent writers can race in the same directory. Prefer a single delegation call that creates isolation, and assign distinct ports or container names to parallel tasks that run servers.

How do I verify work completed by a delegated AI agent?▼

Never trust the agent's Final Report alone - inspect the actual diff and execute what was built. Run the backend against a seeded database, run frontend builds and tests, diff files claimed as copied, and bring up compose stacks for real.

Does this orchestration methodology require a specific tool or protocol?▼

No, the methodology is protocol-agnostic and works whether the Executor is driven over MCP, a CLI wrapper, or manual handoff. Concrete tool names for a specific binding live in that binding's own repository.

Why does a delegated AI task appear stuck while still running?▼

An OS-level WORKING verdict only proves a socket is open, not that the task is converging. Cross-check against cost, token count, and recorded plan - a low CPU-to-elapsed ratio with ballooning cost usually means the task is re-reading its own context and should be cancelled.

What are the limitations of git worktree isolation for parallel tasks?▼

Worktrees do not isolate shared host resources like ports and container names, nor the repository's own .git/config. Parallel tasks need explicitly assigned resources, and tasks touching repo-level configuration must document activation commands instead of running them.