No Negative Echo Advanced: Red-Team & Devil's Advocate Workflow

In the PRD review everyone politely offered 'a few optimizations'; after launch, users voted with their feet. The post-mortem found every fatal flaw was already written in the doc — nobody wanted to be the one to say it out loud. You didn't need a friendlier review; you needed a structured pre-mortem.

Deep Skill Finder community · PM in post-mortem

Being your own devil's advocate is a paradox: humans can't argue against what they sincerely believe. You tried making two AIs debate each other — they agreed within one round. A debate without forced adversarial mechanics is theater.

Deep Skill Finder community · adversarial-testing user

Before the boss signs off, 'any dissenting views?' meets total silence — not for lack of views, but nobody wants to be the party pooper. What you want is a role that fears no offense, laying out conflicting interests, hidden assumptions, and underweighted risks one by one.

Deep Skill Finder community · meeting-room survivor

Lazy path · recommended

Install deep-skill-finder once — let your Agent find Skills for you

Whatever "Build a workflow from red-team review to counterargument generation" task you have, MEYO already has the matching Skills. You don't need to hunt one by one — install deep-skill-finder and your Agent will auto-search MEYO's 50,000+ Skill library on every task, recommending the best match. You just say what you want to do — matching Skills is the Agent's job.

Copy → paste into Claude Code / Codex / Cursor / WorkBuddy / CatPaw · Agent installs automatically:

Please install the deep-skill-finder skill: download the archive from https://www.meyo.life/skill-finder, extract it into your local skill directory and enable it.

Q: How do red-team reviews actually work?

A: Use red-team-review (1,195 installs): it simulates multiple hostile expert roles attacking your plan, strategy, PRD, or release proposal from every angle, and outputs a severity-ranked risk list, a pre-mortem report, and 'the specific improvements that would most strengthen the plan.' For more bite, red-team-pro (634 installs) plays the pickiest opponent, systematically hunting flaws, risks, and counterexamples.

Q: How do I make the AI actually disagree instead of going through the motions?

A: Two mechanisms, both required. Forced opposition: devils-advocate-new (118 installs) generates rebuttals from three fixed angles — conflicting interests, hidden assumptions, black swans — leaving no room for mutual back-patting. Expert debate: karpathy-llm-simulator (107 installs) simulates multiple experts with opposing views arguing with each other, not one model asking itself questions.

Q: Isn't 'question everything, solve nothing' too thin a deliverable?

A: The questioning IS the deliverable: a rebuttal list tells you where you'll be attacked, a risk list tells you where it blows up, and improvement prompts tell you what evidence to gather next. devils-advocate-new states it outright — 'you are the devil's advocate, not the consultant.' Separating problem-finding from problem-solving is exactly what keeps the critique honest.

Q: How do I judge the quality of counterarguments?

A: Three checks: specificity ('the channel assumption breaks in Q4' beats 'there is market risk'), evidence (againstai requires sources and confidence levels on every judgment), and verifiability (red-team-review's pre-mortem report can be checked line by line after the fact).

Q: Do red-team and anti-sycophancy skills conflict?

A: No — layer them. anti-sycophancy is the daily baseline ensuring the AI doesn't flatter by default; mount red-team-review for dedicated review and decision scenarios; run devils-advocate-new as the final opposing check before a major call. The shared foundation is honesty: honest-agent keeps the agent from fabricating evidence to feed you fake counterarguments.

Q: Can the AI's pushback steer me wrong?

A: Yes — if you only ever hear the opposing side with no adjudication. The right pattern is karpathy-llm-simulator's: let opposing views clash fully, then you decide on the complete pro-and-con evidence. The AI lights up blind spots; the judgment call stays with you.