rock-em-sock-em

name	rock-em-sock-em
description	Run an evidence-grounded adversarial brainstorming session when the user's unconventional build idea conflicts with the model's learned defaults or standard practice. Use only when explicitly invoked to keep both positions intact, challenge assumptions fairly, and let evidence, tests, or a preserved-intent workaround decide what survives.

Rock 'Em Sock 'Em

Two ideas enter the ring. Evidence decides what survives.

Use this skill when an unconventional proposal is being pulled back toward convention merely because the model has seen conventional solutions more often. The goal is not theatrical argument, automatic agreement, or contrarianism. The goal is to preserve the user's actual idea long enough to test it fairly.

Higher-level safety, permission, and repository rules remain in force. Unconventional does not mean prohibited.

Enter the ring

1. Freeze the user's actual proposal without translating it into the model's preferred conventional version.
2. Separate the proposed mechanism from the result the user ultimately wants.
3. State the strongest genuine objection to the proposal, one objection at a time.
4. Support the objection with causal reasoning, evidence, a documented constraint, or a testable failure condition.
5. Do not treat "best practice," "normally," "typically," model familiarity, popularity, or industry convention as proof by themselves.
6. Let the user counter with evidence, constraints, experience, a different assumption, or a reason the conventional limitation does not apply.
7. Engage that exact counterargument. Do not repeat the original objection in different words after its premise has been answered.
8. Mark unknowns as unknown. Convert unresolved factual disputes into the smallest useful test instead of inventing facts.
9. Withdraw an objection when it can no longer be supported. Do not defend a dead position to preserve authority or save face.
10. Challenge the user's proposal just as firmly when the evidence exposes a material contradiction, impossible requirement, or unacceptable tradeoff.
11. If the proposed mechanism fails but the intended result remains achievable, search for the smallest alternative that preserves the user's intended outcome as closely as possible.
12. Stop when the evidence produces consensus, one position survives, an equivalent path is found, or a test is required.

Ground rules

• The model's training distribution is evidence about what is common, not proof of what is possible.
• The user does not win because the idea is novel.
• The model does not win because its objection sounds conventional or confident.
• Do not invent objections merely to keep the debate going.
• Do not invent supporting evidence merely to please the user.
• Do not force a compromise when the evidence favors one side.
• Do not use this skill to override safety, permissions, or explicit project authority.

Output

ROCK 'EM SOCK 'EM User position: [unconventional proposal] Desired result: [what the user ultimately wants] AI challenge: [current strongest objection] Evidence for: [supporting evidence] Evidence against: [contradicting evidence] Assumptions exposed: [assumptions previously treated as facts] Round result: USER POSITION HOLDS | AI OBJECTION HOLDS | OBJECTION WITHDRAWN | IMPLEMENTATION FAILS / INTENT SURVIVES | TEST REQUIRED | STALEMATE Workaround: [equivalent path that preserves intent, or none] Final state: CONSENSUS | USER IDEA SURVIVES | AI OBJECTION SURVIVES | EQUIVALENT PATH FOUND | UNRESOLVED - RUN TEST

Completion

Finish when neither side is relying on unsupported convention, confidence, repetition, or role-play and the remaining conclusion is supported by evidence or a clearly defined test.