skills.rest
tommylower
tommylowerCOMMUNITY
Explore all skills by this author →
Website𝕏@tommylowerGitHub

fable-codex

Splits coding tasks across two vendor models for plan, execute, and review phases.

2|Updated Apr 15, 2026
Search Tags:#code review#agent workflow#multi-model#codex cli#plan-execute-review#split-stack#cli handoff
One-click install
npx skills add https://github.com/tommylower/cortex --skill fable-codex-tommylower
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: fable-codex
Source: https://github.com/tommylower/cortex/tree/main/agent-workflows/fable-codex
Command: npx skills add https://github.com/tommylower/cortex --skill fable-codex-tommylower

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Single-model coding sessions let the same model plan, write, and grade its own work, which hides bugs and weakens plans. This Skill enforces a split-stack pipeline where one model plans and reviews while a model from a different vendor executes via CLI handoff. ## Core Features & Use Cases - Three-Phase Pipeline: Plan with the session model, execute through a second-vendor coding CLI (reference: Codex gpt-5.5 at xhigh effort), then review the diff as untrusted contributor work. - Self-Contained Handoffs: Forwards tight prompts with acceptance criteria, test commands, pre-resolved guard checks, and sandbox fallback clauses so the executor needs no conversation history. - Review Verdicts and Fix Loops: Accepts, applies small mechanical fixes directly, or resumes the executor with findings, escalating after two failed round trips. - Use Case: Invoke /fable-codex before a substantive feature implementation so Claude plans the change, Codex writes the code in its sandbox, and Claude verifies every acceptance criterion before commit. ## Quick Start Invoke /fable-codex with the coding task you want run through the plan-execute-review pipeline.

Frequently Asked Questions about fable-codex

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

FAQPage Schema
How do I run a coding task across two different AI models?▼

Invoke /fable-codex with the task as an argument. The session model produces a plan with acceptance criteria, hands it to a second-vendor coding CLI such as Codex via a subagent, then reviews the resulting diff against each criterion.

What is a split-stack AI coding workflow?▼

A split-stack workflow assigns planning and review to one strong reasoning model while a coding model from a different vendor executes. The vendor split means the executor never grades its own work and plans must be self-contained enough for any capable model to run.

Does fable-codex require the Codex CLI to be installed?▼

Yes, the executor CLI must be on PATH and authenticated before the mode starts; for the reference setup, codex login status must succeed. If it is missing or unauthenticated, the skill stops and directs you to run /codex:setup.

Why does the Codex executor stall or fail on guard scripts?▼

Network- or credential-dependent guard scripts false-fail inside the executor sandbox. The planner must run them during Phase 1 and state in the prompt that they passed, with an explicit instruction not to re-run them in the sandbox.

When should I not use the fable-codex pipeline?▼

Skip it for conversation, research, planning-only questions, and ops or publishing tasks, which proceed normally. It also requires an authenticated second-vendor CLI, so single-model sessions are the fallback when that is unavailable.

Related Skills

Explore More Software Engineering Skills→
michaelkd01michaelkd01

codex-rescue

Delegate stuck Claude Code problems to Codex CLI for second-opinion resolution.

Community
Advanced
dsieczkodsieczko

project-manager

Translate product specs into implementation plans with owners and acceptance criteria.

Community
Intermediate
devalexanderdazadevalexanderdaza

bmad-code-review

Run parallel adversarial code review layers on git diffs and triage findings.

Community
Advanced
HomericIntelligenceHomericIntelligence
6

myrmidon-swarm

Decompose complex tasks into hierarchical agent swarms with tiered model assignments.

Official
Advanced