You're in Control

Kyew never changes how your AI behaves without your say-so. Every skill and every custom tool requires your explicit approval before it goes live.

Why This Matters

Without approval, your AI could:

• Learn the wrong lesson from a one-off mistake
• Save a temporary workaround as a permanent process
• Create tools that don't do what you intended
• Apply knowledge you haven't verified

With approval:

• You review everything before it's active
• Only validated knowledge gets applied
• You can reject, edit, or deprecate anything at any time
• Your AI reflects what you actually want, not what it guessed

The Approval Flow

Created  →  You Review  →  Approved (live)
                ↓
            Rejected (with your feedback)

1. Something Gets Created

When your AI generates a skill from your patterns, or builds a custom tool you requested, it starts as a draft. It won't be used until you say so.

2. You Review It

Look at pending items anytime:

"Show me pending skills"
"What tools are waiting for approval?"

Check:

• Does this accurately capture what I want?
• Is the scope right — not too broad, not too narrow?
• Would this actually help me in the future?

3. Approve or Reject

Approve to make it live:

"Approve the client reporting skill"
"Approve that GitHub tool"

Reject if it's not right:

"Reject that skill — the process has changed since then"

4. It's Live

Approved skills are suggested automatically when your AI detects a relevant situation. Approved tools are available whenever you need them.

After Approval

You're never locked in. You can always:

Update — if something needs tweaking:

"Update the reporting skill to include the new Q1 metrics template"

Deprecate — if something is outdated:

"Deprecate the old deployment skill, we changed the process"

Roll back — if an update made things worse:

"Roll back the reporting skill to the previous version"

Custom Tool Approval

Custom tools (especially code runners) go through the same approval process. Since they can execute logic and connect to your services, your AI will always ask you to review and approve before activation.

This means you never accidentally create a tool that does something unexpected.