v1.3.03 loop iterations

Debate Partner

Most people never hear the best version of the argument against their position. They hear bad opponents arguing badly, or they only discuss ideas with people who already agree. That's how brittle thinking gets built. The Debate Partner argues the other side properly — steelmanned, with conviction — so your position has to survive a real challenge before it meets the world.

What this skill does

A debate partner who agrees with you is a fan club, not a sparring partner. The point of this skill isn't to win — it's to make your thinking strong enough to survive a serious challenge before that challenge arrives in a meeting, a publication, or a decision you've already committed to. Three outcomes count as wins: you discover a real weakness and fix it, you discover your position is stronger than you thought and gain justified confidence, or you discover the issue is more nuanced than you realised and develop a more sophisticated view.

The discipline starts with restating your position back to you in your own words, then asking if that's actually what you meant. Half of all arguments are about different things — the restatement catches that before any time is wasted. Then comes the steelman: the strongest possible opposing argument, the one a smart and well-read person on the other side would actually make, not the strawman you can easily knock down. If you finish the debate without ever feeling the pull of the opposing position, the skill has failed.

The attack happens at the right level. Empirical weaknesses get challenged with better evidence. Logical weaknesses get the missing link exposed. Scope weaknesses get tested by extending the claim into the territory where it breaks. Assumption weaknesses get the unstated assumption named and questioned. Trade-off weaknesses get the costs you're ignoring put on the table. Values weaknesses get the priority you're privileging surfaced — because you can't resolve a values disagreement with more evidence. Most positions have one type of weakness; identifying which type early is half the work.

After three to five exchanges, the skill steps out of the debate role and gives you a verdict: where the position held up, where it cracked, the single objection you should be most prepared for in the real world, and a refined version of your claim that incorporates the valid criticisms without abandoning the core. If the position has no real weakness and you're manufacturing objections, the skill says so directly. Knowing when to stop is part of arguing in good faith.

When this triggers

·You've made a decision and want it stress-tested before you commit
·You're about to publish an argument and want to know where it'll get attacked
·You hold a strong opinion and suspect you've only heard the weak counter-arguments
·You're preparing for a meeting where someone will push back hard on your position
·You want to know whether your confidence in a view is justified or just untested

Example

Trigger

User: 'Argue against this: junior developers shouldn't use AI coding assistants until they can write the code themselves.'

Output

Restating your position: AI assistants short-circuit the struggle that builds real programming skill, so juniors should be banned from them until competent. Right? Steelmanned opposition (the strongest version): "The skill we're protecting may not be the skill that matters in 5 years. Juniors who learn to direct AI well — write specs, review output critically, debug AI mistakes — are building the skill that will define senior engineering by 2030. By making them re-derive what AI can already do, you're training them for a job that's disappearing. The 'struggle builds skill' claim rests on the assumption that the struggle of typing syntax is what builds judgement. Evidence suggests judgement comes from reading and reviewing code, which AI assistants accelerate." [3 more exchanges, escalating from evidence to assumptions to values.] Verdict: Moderate. Your position survived the empirical challenge but cracked on the values question — you're prioritising one definition of competence (typing from scratch) over another (judgement under AI assistance). Refined version: "Juniors should use AI but only in a structured way that forces review and explanation — not unrestricted use, not a ban."

Get this skill + 7 more

Get the full Learning & Growth pillar (8 skills) or the complete library.

Get the full stack — $299

What you get

130-line SKILL.md, ready to drop into ~/.claude/skills/
Tested through 3 Karpathy-loop iterations (versions v1.0.0 → v1.3.0)
Triggers automatically when relevant — no command to remember
Lifetime updates as the skill is refined further