CEOAI GitHub

A thinking partner for your next move

Not surewhat’s next?Ask your CEO.

Talk through the idea. Find the weak assumptions. Turn a considered decision into work you can check.

For Codex, Claude Code & Cursor

Built-in security checks

How the work moves

1 / 3 · Analysis: clarify the goal

Workflow illustration

CEO AI / 01

A useful question beats a rushed decision.

CEO helps work out what is worth doing, then coordinates the next steps. Other AI assistants join when your tool supports them.

Fictional dialogue · a little satire
You

Make it sparkle! Put the API key on the site too. Easier setup.

CEO

That makes your budget public too. Keep secret keys on the server. What should the product help people do?

CEO

Start with booking and confirmation. Test access controls and the user flow; make claims the evidence supports.

Sparkle can wait. Booking should work.

Read the full dialogue ↗

CEO AI / 01 / 03

From “what if” to a checked result.

01

Think it through

Separate the outcome you want from the proposed solution. Check that hypothesis against evidence, primary sources and applicable established practices before recommending an approach.

02

Make a plan

Break work into clear steps. Bring in other AI assistants when your tool supports them.

03

Check the outcome

Tests for code, contradictions in writing, missing steps in a plan. State what remains uncertain.

CEO AI / Built-in security

Security checks start
at indexing.

Check selected skills and agents locally. CEO AI looks for risk signals in approved directories without running their content.

  1. Install CEO AI.

    Start a new chat and select the core skill. Installation does not start a scan.

  2. “Check my skills.”

    Use the ready-made request. Confirm the directories and private report location first.

  3. Read findings and gaps.

    Review the signals, what was not checked and the suggested next action.

Example statuses · not a scan report
No findings
Check origin and permissionsThe rules found no risk signals.
Review required
Inspect the flagged actionThe signal may be a false positive.
Incomplete
Read the coverage gapsSome content was not analyzed.

Trust stays unverified. No findings does not prove safety; a report does not grant execution permission.

CEO AI / Memory

Keep the decisions that worked.

Approved patterns and verified lessons can help the next task. You decide what CEO AI remembers.

MEMORY OFF BY DEFAULT

  1. Check the evidence
  2. Save an approved pattern
  3. Reuse when it fits

Illustrative project pattern

Graphite / warm accent

SurfaceAccentText
StatesMobile layoutsKeyboard access

Check what exists first.

Before a new design, CEO checks existing components and suitable documented practices. It compares meaningful options with sources, then explains whether to reuse or build.

Your memory. Your controls.

Project lessons and user preferences stay in separate scopes. CEO announces each save. Ask to view, correct, delete or disable memory. Recalled records may enter your AI app’s model context.

“Show my CEO AI memory”

Memory and templates ↗

CEO AI / Your kind of work

Same questions. Different professions.

CEO started with software work. The approach also helps with a brief, a proposal or a research question: clarify the goal, challenge the plan, do the work and check it.

01DevelopmentFix the cause before rewriting

Your task

“Reports take too long. Should we rewrite the service?”

How CEO helps

Checks measurements, compares a targeted fix with a rewrite, and plans implementation and tests.

What you can produce

A measured change, a test report and a decision about the rewrite.

02MarketingTurn an idea into a test

Your task

“We have a new product and a limited launch budget.”

How CEO helps

Clarifies the audience and offer, challenges channel assumptions, drafts campaign messages and a small experiment.

What you can produce

A campaign brief, message drafts and a plan for measuring the test.

03SalesPrepare a relevant proposal

Your task

“Help me prepare for a meeting with a potential customer.”

How CEO helps

Uses your notes to identify the customer’s problem, unanswered questions and unsupported promises in the offer.

What you can produce

A meeting agenda, discovery questions and a proposal draft to review.

04ManagementChoose what the team should do first

Your task

“Three projects matter, but the team can only take on one.”

How CEO helps

Compares goals, effort, dependencies and the cost of delay. Challenges priorities and asks what would change the decision.

What you can produce

A decision note, a sequence of work and checkpoints for reconsidering it.

05DesignCheck a brief before drawing screens

Your task

“Users abandon this flow. Help me decide what to change.”

How CEO helps

Reviews the available evidence, separates usability problems from assumptions and compares design options.

What you can produce

A revised brief, user-flow options and a usability test plan.

06ResearchFind what the evidence really supports

Your task

“These sources disagree. Help me make sense of them.”

How CEO helps

Compares the sources you provide, separates findings from assumptions and flags missing evidence. Web research needs browsing tools.

What you can produce

A comparison with source references, open questions and a next research step.

These are example scenarios, not customer success stories. Results depend on the model, the materials you provide and the tools your app can access. CEO does not automatically gain access to a CRM, ad account or design tool.

CEO AI / v0.6.0

One message.
A place to start.

Open your project in your AI tool. Copy this message and send it to your agent. It will inspect the instructions and install the skill.

For Windows, macOS and Linux. The agent checks whether your app supports plugins; the project installer remains available.

Start a new chat after installation. Plugin: ceo-ai:ceo-ai (Codex), /ceo-ai:ceo-ai (Claude Code). Project skill: $ceo-ai (Codex), /ceo-ai (Claude Code, Cursor).

Plugins use your app’s catalog and cache, including user-level settings. The project installer stays inside the project.

Plugin setup and migration ↗
Already use Microsoft APM?APM instructions ↗

CEO AI / Q&A

A few things to know.

CEO AI is under active development. We test and refine the orchestration skill based on test results and feedback.

Open the help guide ↗
What does it install?

One skill with its supporting instructions. No new model, accounts or automatic registration of other agents.

Do I need Docker?

No Docker is needed for the builtin check. The optional CEO AI Security module requires it. See the separate Docker walkthrough for the supported platform and setup.

Will it just agree with me?

CEO checks whether your proposed solution serves your goal. Recommendations should follow evidence, authoritative primary sources and applicable established practices, with gaps stated clearly. You decide; neither agreement nor disagreement is automatic.

What has been tested?

Native plugin installation, updates, removal and skill discovery were checked on macOS. The project installer is checked in Windows, Linux and macOS CI. Model invocation remains unverified. See the plugin guide for versions and limits.

What if installation stops?

Keep your files. Ask the agent to explain the conflict or missing permission. The help guide covers missing skills, interrupted updates and older installations.

AI can be wrong. A second AI review can be wrong too. CEO gains no extra permissions and does not replace your project’s required security reviews.