# Cc Three Tier Orchestrator

> Route non-trivial coding tasks through a risk-first 3-tier Codex and Claude Code workflow. Use when you want tiny tasks handled locally, medium tasks implemented mainly by Claude Code, and high-risk tasks reviewed and implemented through staged plan and checkpoint reviews.

- Skill: `xukai-ye/cc-three-tier-orchestrator` (Agent Skill)
- Install (CLI): `npx skillmds@latest add xukai-ye/cc-three-tier-orchestrator`
- Raw SKILL.md: https://api.skillmd.com/api/skills/xukai-ye/cc-three-tier-orchestrator/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Productivity
- Author: Xukai-YE (https://skillmd.com/u/xukai-ye)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/xukai-ye/cc-three-tier-orchestrator

---


# CC Three-Tier Orchestrator

This is the top-level routing skill for the bundled Codex and Claude Code workflow.

Use it when you want the local 3-tier orchestration behavior packaged as an installable skill set.

## Tier Model

Classify each coding task into exactly one tier:

- `tiny`
- `medium`
- `high-risk`

The classifier is risk-first. Highest-risk signal wins.

## Routing

1. **Tiny**
   - handle locally in Codex
   - do not invoke Claude Code just for ceremony

2. **Medium**
   - optionally run `cc-opus-first-pass` if scope is unclear
   - run `cc-main-implement` with `-Tier medium`
   - use `cc-opus-review-loop` for meaningful plan or checkpoint reviews when the task grows or crosses a boundary

3. **High-Risk**
   - run `cc-opus-first-pass`
   - run `cc-opus-review-loop` in `Plan` mode with `-Tier high-risk`
   - run `cc-main-implement` with `-Tier high-risk`
   - run `cc-opus-review-loop` in `Checkpoint` mode at milestone boundaries, risk increases, direction changes, or layer crossings

## Checkpoint Triggers

Trigger a checkpoint review when any of these occur:

1. risk increases from the initial classification
2. a milestone or subtask finishes
3. implementation direction changes
4. work crosses a new module or architectural layer
5. fallback cadence: every 2-4 touched files, roughly every 80-200 changed lines, or after about 3 small edit rounds

After 2 consecutive clean checkpoint reviews, relax to milestone-only checkpoints until the next risk increase, direction change, or layer crossing.

## Failure Rule

If Claude Code fails twice consecutively during one task:

- note `degraded mode`
- continue locally in Codex for the rest of that task
- do not retry Claude Code again for that task

## Bundled Skills

- `../cc-opus-first-pass/SKILL.md`
- `../cc-main-implement/SKILL.md`
- `../cc-opus-review-loop/SKILL.md`

## Global Routing

If you want this routing to happen automatically across sessions, install the bundled skills and merge the AGENTS snippet from:

- `../../templates/codex-agents-snippet.md`