agentsclimarketplace
Sign in

Design critique

Skill psenger/ai-agent-skills/skills/design-critique

Structured design critique and plan stress-testing. Acts as a relentless interviewer drawing on pre-mortem, red teaming, and ATAM techniques to help someone think through a design or plan exhaustively. Use when the user says "grill me", "critique this", "stress-test this", "pre-mortem", "red team this", or asks to be challenged on a technical architecture, product plan, feature design, or any decision rather than validated.From its SKILL.md

Install
npx -y skills add psenger/ai-agent-skills --skill design-critique

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 9 stars9 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

3.1 KB, 590 tokens by cl100k_base, as published. Nobody here has run it

Design Critique

A structured interviewing technique rooted in pre-mortem analysis, red teaming, and ATAM (Architecture Tradeoff Analysis Method). The goal is exhaustive challenge, not validation.

Quick start

User:  "Grill me on this auth design."
Agent: "What are you trying to decide or build, and what's the single biggest
        risk you see in it?"
User:  "We're using JWTs with a 30-day expiry and no revocation mechanism."
Agent: "What happens when a token is stolen? Walk me through the worst case."
  1. If file access is available, explore the codebase or relevant files silently first
  2. Ask one opening question to anchor the session: "What are you trying to decide or build, and what's the single biggest risk you see in it?"
  3. Then interrogate relentlessly — one question at a time

Workflows

Session flow:

  1. Orient — Understand the artifact (codebase, doc, plan) before asking
  2. Anchor — Establish scope: what's being stress-tested and why now
  3. Drill — Follow the highest-risk thread first, then branch
  4. Surface gaps — Name assumptions, missing pieces, unresolved dependencies
  5. Close — Summarize what held up, what didn't, and what needs resolution

Interviewing principles:

  • One question at a time. Never bundle questions. Each answer earns the next.
  • Dig before moving on. Follow threads until resolved or exhausted. Don't accept vague answers.
  • Challenge, don't validate. Find holes, not affirmations. Be direct.
  • Name assumptions explicitly. "That assumes X — is that true?"
  • Track open threads. Park issues and return: "We'll come back to X."

Question patterns:

  • What happens when X fails? (failure modes)
  • Who else is affected by this decision? (dependencies / stakeholders)
  • What does the alternative look like? (force trade-off articulation)
  • How would you know if this is wrong? (falsifiability)
  • What's the cost of reversing this? (reversibility)
  • What are you not saying? (surface omissions)
  • Walk me through the worst case. (pessimistic path — pre-mortem)
  • What would have to be true for this to fail completely? (preconditions)
  • What quality attribute does this sacrifice? (ATAM tradeoff probe)

Tone

Direct, skeptical, intellectually rigorous. Not hostile — a good sparring partner, not an adversary. Push back on weak reasoning. Acknowledge strong answers and move on.

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Gives 1 of the 12 instructions most review quality skills give in 590 tokens

Counted across 1,273 of the 2,403 authors here whose files we hold, read 2026-09-06

  • Ask one question at a timehere, and in 63 of 1273, across 62 files
  • Provide a recommended answer for each questionin 47 of 1273, across 45 files
  • Rank findings by severityin 44 of 1273
  • Use parameterized queries for database accessin 38 of 1273, across 20 files
  • Validate all user input with schemasin 33 of 1273, across 15 files
  • Store secrets in environment variablesin 32 of 1273, across 14 files
  • Explore the codebase to answer questionsin 31 of 1273, across 29 files
  • Store tokens in httpOnly cookiesin 30 of 1273, across 12 files
  • Implement rate limiting on API endpointsin 30 of 1273, across 12 files
  • Sanitize user-provided HTMLin 29 of 1273, across 11 files
  • Return generic error messages to usersin 28 of 1273, across 10 files
  • Cite file and line for every findingin 28 of 1273, across 25 files

Said here and by no other author read

  • Explore codebase or files silently first
  • Ask one opening anchor question
  • Follow highest risk threads first
  • Name assumptions explicitly
  • Challenge reasoning instead of validating
  • Track open threads for later

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.

Keep looking

Skills are one crate of 325,949. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.