Update skill creation and modification guidance to use gates

[nathanjmcdougall]

See the information from https://blog.fsck.com/2026/04/07/rules-and-gates/ to get the general idea:

The difference: a rule has an opt-out path (I can rationalize "I'll do it after this one thing"). A gate doesn't — the next action is blocked until the gate condition is met. Both are now written as gates with an explicit sequence: thing happens → gate condition → then proceed.

Here's a concrete example from an agent today:

Before (rule): "Verify claims with web research before asserting them."

That's a rule. It lives in my head. When I was answering your question about whether audience assessment books exist, the rule didn't fire — I synthesized from training data, it felt confident, I stated it as a finding. The rule had an invisible opt-out: this feels solid enough, I'll skip the search.

After (gate): "When a claim about what exists or doesn't exist is forming → web search happens → URLs in hand → then I speak."

The gate makes the sequence explicit and gives me something to check: do I have URLs? If no, I haven't cleared the gate. "I think based on my training data" is the only thing I'm allowed to say until I do. There's no version where I proceed without either doing the search or flagging the claim as unverified.

The practical test of whether something is a gate rather than a rule: when I'm about to skip it, does the gate formulation give me a concrete question I can't answer? "Do I have URLs?" is concrete. "Did I verify this?" is too easy to answer yes to without having actually done anything.

We should always use gates, and we should update all existing skills to use them.