Sponsored by
Which coding agent do you use more?
WTF is the shared-folder setup?: Your code, docs, notes, scripts, and references live in one normal project folder.
Claude Code and Codex can both work from that same folder if you give them the right instruction files.
People treat coding agents like football teams.
Claude Code vs Codex.
Cursor vs Opencode.
This model vs that model.
But that is not the useful way to think about it.
Sometimes Claude Code gets stuck.
Sometimes Codex gets stuck.
Sometimes one agent loops on the same error for ten minutes, and another fixes it in one pass.
So the real move is not:
Which agent is the best?
The better move is:
Can my project survive switching agents?
That is the shortcut.
I am Alex, welcome to ShortCu8 by Innov8.
Lets Dive Deep 🐰
⭐Today's Shortcut
Do not lock your project inside one AI tool.
Make your project portable.
Use one folder that both Claude Code and Codex can understand.
Then when one agent gets stuck, hand the session to the other.
Simple idea:
Claude Code stuck?
Make a handoff summary.
Open Codex in the same folder.
Paste the summary.
Keep going.Same works the other way too.
Codex stuck?
Hand it to Claude Code.
The goal is not loyalty.
The goal is shipping.
The Three Layers
Every coding-agent project has three layers.
1. Shared knowledge
This is the normal project stuff:
code
docs
scripts
decisions
references
brand assets
notes
markdown files
Any agent can read this.
This layer should not belong to one tool.
2. Workflows and skills
These are the repeatable recipes:
review this dashboard
run this test
create this report
pull this data
write this style of post
Claude Code and Codex both use markdown-style skills.
The folder location changes, but the idea is the same.
3. Agent config
This is where the tools differ.
Claude Code has its own files.
Codex has its own files.
Same job.
Different wrapper.
Claude Code vs Codex Folder Setup
Claude Code usually looks for:
CLAUDE.md.claude/.claude/skills/
Codex usually looks for:
AGENTS.md.codex/.agents/
Think of CLAUDE.md and AGENTS.md as the onboarding document.
They tell the agent:
what the project is
what matters
how to work
what to avoid
how to verify changes
If you already have CLAUDE.md, ask Codex to convert it into AGENTS.md.
Prompt:
Read CLAUDE.md and create an AGENTS.md for Codex.
Keep the same project context, rules, commands, and verification steps.
Adapt only the parts that are specific to Claude Code.That gives Codex the same project brain.
Skills Are Mostly Portable
This is the nice part.
Claude Code skills are markdown files with YAML frontmatter.
Codex skills are also markdown files with YAML frontmatter.
Same basic format.
Different folder.
Claude Code checks:
.claude/skills/
Codex checks:
.agents/
So if you have a useful skill, copy it into both places.
Example:
newsletter-writer.md
bug-handoff.md
security-check.md
No need to rewrite the whole thing from zero.
Just make sure any tool-specific commands are updated.
The Session Handoff Prompt
This is the most useful part of the whole setup.
When one agent is stuck, ask for a handoff summary.
Use:
Create a session handoff summary for another coding agent.
Include:
- goal of the task
- files changed
- commands run
- errors seen
- decisions made
- what worked
- what did not work
- what to try nextThen open the other agent in the same project folder.
Paste the handoff.
Ask:
Continue from this handoff.
Do not repeat failed attempts.
Inspect the current files first.
Then propose the next fix.That one move can save a stuck session.
Simple Setup
If you want the 5-minute version:
Open your Claude Code project folder.
Create AGENTS.md from CLAUDE.md.
Copy useful skills from .claude/skills/ into .agents/.
Keep shared docs in normal markdown.
Open Codex in the same folder.
Ask Codex to inspect the project and confirm it understands the setup.
When either agent gets stuck, use the handoff prompt.
That is it.
You now have a two-agent project.
Now go build something great.
The ShortList
🛠️Cool Tools of the Week:
Capcut with Gemini: The popular video editing tool is partnering with Google to allow users to create and edit images in the Gemini app.
LoRAs for Krea 2: A new fine-tuning system to train the company’s foundation AI image model to specific styles.
ElevenLabs Speech Engine: Turn your existing chat agents into full voice agents with natural language prompts.
Cohere Command A+: The AI lab’s “most powerful LLM yet,” open source and optimized to run on as little hardware as possible.
📩 Innathe Shortcu8 engane undarunnu 👇️?
We read every reply - just reply to this email and let us know how we can improve !
Appo adutha Shortcu8il kanaam bie…👋
If you read till here, you might find this interesting
#AD1
The World's Biggest Dev Event Hits Silicon Valley
WeAreDevelopers World Congress comes to San José, CA — September 23–25, 2026. 10,000+ developers, 500+ speakers, and the full software development lifecycle under one roof, in the heart of Silicon Valley.
Kelsey Hightower. Thomas Dohmke (fmr. CEO, GitHub). Christine Yen (CEO, Honeycomb). Mathias Biilmann (CEO, Netlify). Olivier Pomel (CEO, Datadog). The people actually building the tools you use every day — all on one stage.
AI, cloud, DevOps, security, architecture, and everything real builders ship with. Workshops, masterclasses, and the official congress party.
#AD2
Distinguish real intent from malicious intent.
hCaptcha User Journeys finds malicious intent across sessions, devices, and apps. Detect intent signals that expose risk before it escalates.
Understand motives, not just outcomes. Book a demo and find out how it works.