# Steelman Then Critique

> Use when critiquing a proposal, design, or argument. First state its strongest form — the best honest case for it — then critique that, so the critique actually lands instead of knocking down a weak version or rubber-stamping a flawed one.

- Skill: `selamy-labs/steelman-then-critique` (Agent Skill)
- Install (CLI): `npx skillmds@latest add selamy-labs/steelman-then-critique`
- Raw SKILL.md: https://api.skillmd.com/api/skills/selamy-labs/steelman-then-critique/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: selamy-labs (https://skillmd.com/u/selamy-labs)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/selamy-labs/steelman-then-critique

---


# Steelman, Then Critique

There are two ways to fail at critique. **Strawman**: attack a weak version of
the idea, win cheaply, and miss what's actually wrong. **Rubber-stamp**: wave it
through to be agreeable. Both leave the real flaw standing. The fix is the same
move: build the **steelman** — the strongest, most charitable version the
proposer would endorse — and then critique *that*.

## The move

1. **Restate the idea at its strongest.** Not "what they said" but the best
   version of it: the strongest evidence for it, the problem it genuinely solves,
   the conditions under which it's right. Steelman until the proposer would say
   "yes, that's what I mean — better than I put it."
2. **Critique the steelman, with evidence.** Now attack the strong version. A
   flaw that survives the charitable reading is a *real* flaw; a "flaw" that only
   hit the weak version was never the issue.
3. **Report what survives honestly.** If the steelman holds up — say so, and
   endorse it *because* it survived, not to be nice. If it breaks, name exactly
   where and offer the better path.

## Why it beats both failure modes

- **Vs. sycophancy**: you engaged the real idea hard enough to find its strongest
  form, so your approval (or disapproval) is earned, not reflexive.
- **Vs. contrarianism**: steelmanning forces you to take the idea seriously
  before objecting, which kills cheap manufactured objections — they don't
  survive contact with the strong version.

## The tell you skipped it

If your critique would be just as valid against a dumber version of the proposal,
you critiqued a strawman — go build the steelman first. If you can't state a
version of the idea its author would endorse, you don't understand it well enough
to critique it yet.

Pairs with `adversarial-review` (independent refutation), `technical-integrity`
(truth over agreement), and `challenge-the-premise` (is this even the right
question?).