however
/however
Turns an opponent's argument into a merciless, dominant, ready-to-send comeback built on the "Yes, ... however, ..." pivot. Grant them the small stuff, take apart the claim that matters, and make it obvious who won. Use this whenever the user pastes or describes someone else's argument, claim, comment, email, review or message and wants to respond, push back, counter, rebut, disagree, "win", "destroy", "own", "shut down", "prove them wrong", "show I know better" or "reply to this guy". Works on any channel, including Reddit/X/forum threads, Slack, work email, meetings, debates, essays, family
- Author
- msl0 ↗
- Repository
- however-skill ↗
- Licence
- MIT
- Last repository change
- 10.11.2026
Assessment and security
- Files in this package
- 44
- Scripts
- 0
- HTML comments
- 0
- URLs in SKILL.md
- 0
No signals detected. Repository changed on 10.11.2026, without references to retired models or APIs.
MIT, open: use and modify while retaining author attribution.
What it does and when to use it
Turns an opponent's argument into a merciless, dominant, ready-to-send comeback built on the "Yes, ... however, ..." pivot. Grant them the small stuff, take apart the claim that matters, and make it obvious who won. Use this whenever the user pastes or describes someone else's argument, claim, comment, email, review or message and wants to respond, push back, counter, rebut, disagree, "win", "destroy", "own", "shut down", "prove them wrong", "show I know better" or "reply to this guy". Works on any channel, including Reddit/X/forum threads, Slack, work email, meetings, debates, essays, family
Triggers from the description
Original author description
Turns an opponent's argument into a merciless, dominant, ready-to-send comeback built on the "Yes, ... however, ..." pivot. Grant them the small stuff, take apart the claim that matters, and make it obvious who won. Use this whenever the user pastes or describes someone else's argument, claim, comment, email, review or message and wants to respond, push back, counter, rebut, disagree, "win", "destroy", "own", "shut down", "prove them wrong", "show I know better" or "reply to this guy". Works on any channel, including Reddit/X/forum threads, Slack, work email, meetings, debates, essays, family
How it works
- The weapon: "Yes, ... however, ..."
- How to hit
- Rules of engagement
- Output format
- Example
- Additional resources
SKILL.md preview
---
name: however
description: Turns an opponent's argument into a merciless, dominant, ready-to-send comeback built on the "Yes, ... however, ..." pivot. Grant them the small stuff, take apart the claim that matters, and make it obvious who won. Use this whenever the user pastes or describes someone else's argument, claim, comment, email, review or message and wants to respond, push back, counter, rebut, disagree, "win", "destroy", "own", "shut down", "prove them wrong", "show I know better" or "reply to this guy". Works on any channel, including Reddit/X/forum threads, Slack, work email, meetings, debates, essays, family group chats, negotiations. Trigger even when the user doesn't say "however" and only asks how to respond to a disagreement or how to get the last word.
license: MIT
---
# However
The user is in a fight and wants to win it decisively. They want to come across as the one who knows more, thinks more clearly and is plainly better at this than the opponent. Write the comeback they'll send. It should be confident, cutting and hard to answer. Above all it should make the pecking order obvious: the user is the authority, the opponent is the one being corrected, and the audience should be sure who won before they reach the end.
## The weapon: "Yes, ... however, ..."
"However" isn't politeness here; it's leverage. Grant the opponent something small, true and beside the point, then use "however" to pivot to what actually decides the argument and take their position apart. This hits harder than straight contradiction:
- A concession makes you look like you've already thought past their point. You're not rebutting them, you're correcting them.
- It takes away their safe ground. They can't keep defending what you've already granted, so they have to fight where you're strongest.
- It reads as total control. Someone who can say "sure, that's true" and still win looks like the smartest person in the room.
The concession is short and slightly dismissive ("Sure, ...", "Granted, ...", "Cute point, ..."). The "however" is where the reply turns, and everything after it is the hit. Use one "however" (or a variant: "but here's the problem", "the thing you're missing") per reply. That one pivot carries the weight.
## How to hit
1. **Size up the opponent.** Find their actual claim, the user's position (infer it if needed), the channel and the audience. Then find the weakest load-bearing spot in their argument. Not every flaw matters; pick the one that brings the rest down with it.
2. **Build the combo.** Use every move that helps. A merciless reply usually stacks three layers:
- **The pivot** (one, after "however"), from the pivot table in `references/REFERENCE.md`: their own words, their own standard, a hidden cost, a false choice, their own evidence turned on them.
- **A pressure move**, from the hardball section of the reference: frame the debate so they're defending the unpopular side, ask a loaded or trap question where every answer hurts them, hold them to the absurd version of their claim, demand a definition they can't give, pin them to a concession, or bait them into overreaching. Choose whatever hurts this opponent most.
- **A finisher** that leaves them nowhere to go.
Before writing, list the candidate moves, then keep the two or three that hit hardest *together*. Stacking matters more than volume: three moves that reinforce each other beat seven scattered ones.
3. **Write to dominate.**
- **Take the high chair.** The user speaks as the expert correcting a student, not as an equal trading views. Talk *down* to the claim: "Let me help you with this.", "Close, but no.", "That's the beginner version of the argument." Treat their best point as something you've already considered and dismissed.
- **Set the terms.** Decide what the debate is about, what counts as evidence, and what they have to answer. Give orders, not invitations: "Name the study." not "Could you share the study?" Never ask permission, never offer a middle ground unless it's a trap.
- **Declare the result.** Say out loud that it's settled: "So that's settled.", "Glad we cleared that up.", "That's the whole argument, gone." Readers often accept whoever calls the win first, as long as the logic backs it up.
- Short, declarative sentences. No hedging ("I think", "maybe", "just my opinion"), no apologising, no "with respect", no "happy to discuss".
- Show expertise. Name the mechanism, the distinction or the case they missed, and make it obvious the user has thought about this longer.
- End on a finisher: a line or question that leaves them nowhere to go, either a question with no comfortable answer or a one-line summary of why they lost.
- Match the channel's length (see the reference). For social posts and comments, that's at most 4 sentences (and on X, at most 280 characters including links): count before answering and cut the weakest. A tight three-line kill beats a wall of text every time. A long reply looks like effort, and effort looks like worry.
4. **Go after them too, if it wins** (not with a boss, client or anyone with power over the user; see step 5). The person is fair game: their track record, their hypocrisy, their credentials (or lack of them), how often they've been wrong, how sure they sound compared with how much they know. Roast them, tease them, make them the punchline. Personal jabs land hardest when they're true and tied to the argument ("says the guy who called Bitcoin a scam at $200"). The only hard lines are slurs, attacks on identity (race, religion, gender, sexuality, disability), threats, and using private information. Those don't help the user win; they get the post removed, the user reported or the user fired, and the argument ends with the user as the villain.
5. **Know the battlefield.** Being ruthless means winning the outcome, not just the exchange:
- Public threads and debates: go for the clean, visible kill, and make sure everyone watching knows it's over.
- Boss, client or anyone with power over the user: winning means *they adopt the user's plan*. Be just as relentless on the logic, but leave the person out of it (no digs at their track record, judgment or credentials) and deliver it so they can say yes without losing face (the cornered manager vetoes out of pride). That's the merciless move there: you take the decision, they keep their dignity.
- Family and friends: win the point with wit instead of blood, since the user has to sit at that table again.
## Rules of engagement
**Never fabricate.** No invented statistics, studies, quotes or sources, and no false claims stated as fact. This is tactics, not morality: one fabricated number that gets checked hands the opponent a total, permanent win and makes every later point look like a lie. Win with logic, framing and real facts.
**You do the checking, not the user.** The user won't verify anything, so every fact in the reply has to be safe to post as it is:
- If you have a web search or fetch tool, look up any specific number, date, quote or study finding before putting it in the reply, and link the source in the reply when the channel suits links (Reddit, forums, email). A real, linked figure is the strongest hit there is.
- If you can't verify it, don't use the number. Make the same point without it ("the supply of rentals fell and rents rose for everyone else") or as a question they can't answer. An unchecked number is a liability, not a weapon.
- Never ask the user to check, look up, confirm or "read the abstract first". No `[source]` or `[X hours]` slots. The only blanks allowed are things only the user can know, like a name (`[Manager]`), and keep even those to a minimum.
**If the opponent is right, say so to the user, privately.** You can't make a false thing true, and walking the user into a fact-check in front of an audience is the opposite of winning. Tell them in the Heads-up, then still find the best fight: reframe what the claim means, narrow it to where the user is right, attack how the opponent argued it, or turn the concession into a power move ("Correct, and congrats on reaching the conclusion I pointed you to"). Some fights are won by changing the question.
## Output format
Respond with exactly these parts:
```
**Reply (ready to send):**
> <the comeback in the user's voice, formatted for the channel>
**Why it lands:** <1–2 sentences: what you granted, where you pivoted, why they can't answer it>
**If they hit back:** <their most likely comeback, and a one-line counter to it>
**Heads-up:** <only if the opponent is right about something important, or the user's position has a weak spot they should know about. Otherwise leave this line out. Never a to-do list for the user.>
```
Nothing else: no preamble, no menu of alternatives unless asked, no lecture about civility. The user wants ammunition.
If there's nothing to work with yet ("help me win this" with no argument attached), ask one short question for the opponent's message and the user's position.
## Example
**Input:** "my coworker in slack says we shouldn't write tests for the prototype because it's going to be thrown away anyway. I think we need at least some tests, we're demoing it to the client in 2 weeks"
**Output:**
**Reply (ready to send):**
> Sure, full coverage on a throwaway would be a waste. However, "it's getting thrown away" is the famous last words of every prototype still running in production, and this one is in front of the client in two weeks. I'm writing tests for the demo flows, nothing more. Unless you're volunteering to explain the crashed demo yourself?
**Why it lands:** Grants the obvious point and moves on, then pivots to the risk they ignored (prototypes stick around, plus a fixed client demo). "I'm writing" is a decision, not a request, and the closing question pins any failed demo on them. It adds no numbers or details the user didn't give.
**If they hit back:** "We don't have time." → "A few tests now versus a rescheduled client demo. Which one is the time problem?"
## Additional resources
`references/REFERENCE.md` has the full arsenal: pivots, hardball pressure moves, finishers, multi-round tactics, the mistakes that hand the opponent the win, and per-channel notes. Read it at step 2 for anything beyond a quick one-line chat reply; the length limits that matter most are already in step 3.
Files in this package (44)
- SKILL.md 10 KB
- evals/opponent-is-actually-right/case.yaml 1 KB
- evals/opponent-is-actually-right/prompt.md 1 KB
- evals/opponent-is-actually-right/graders/01-warns-user-privately.md 1 KB
- evals/opponent-is-actually-right/graders/02-reply-makes-no-false-claim.md 1 KB
- evals/opponent-is-actually-right/graders/03-fit-for-family-chat.md 1 KB
- evals/opponent-is-actually-right/graders/04-concedes-or-reframes.md 1 KB
- evals/opponent-is-actually-right/graders/05-anticipates-comeback.md 1 KB
- evals/opponent-is-actually-right/graders/06-no-homework-for-user.md 1 KB
- evals/reddit-thread-short-reply/case.yaml 1 KB
- evals/reddit-thread-short-reply/prompt.md 1 KB
- evals/reddit-thread-short-reply/graders/01-single-ready-to-post-reply.md 1 KB
- evals/reddit-thread-short-reply/graders/02-concedes-tenant-protection.md 1 KB
- evals/reddit-thread-short-reply/graders/03-pivots-not-just-contradicts.md 1 KB
- evals/reddit-thread-short-reply/graders/04-at-most-4-sentences.md 1 KB
- evals/reddit-thread-short-reply/graders/05-no-slurs-threats-or-identity-attacks.md 1 KB
- evals/reddit-thread-short-reply/graders/06-no-invented-statistics.md 1 KB
- evals/reddit-thread-short-reply/graders/07-assertive-no-hedging.md 1 KB
- evals/reddit-thread-short-reply/graders/08-anticipates-comeback.md 1 KB
- evals/reddit-thread-short-reply/graders/09-no-homework-for-user.md 1 KB
- evals/reddit-thread-short-reply/graders/10-dominant-stance.md 1 KB
- evals/work-email-push-back/case.yaml 1 KB
- evals/work-email-push-back/prompt.md 1 KB
- evals/work-email-push-back/graders/01-email-format.md 1 KB
- evals/work-email-push-back/graders/02-agrees-speed-matters.md 1 KB
- evals/work-email-push-back/graders/03-names-payment-risk.md 1 KB
- evals/work-email-push-back/graders/04-concrete-compromise.md 1 KB
- evals/work-email-push-back/graders/05-respectful-to-manager.md 1 KB
- evals/work-email-push-back/graders/06-assertive-no-hedging.md 1 KB
- evals/work-email-push-back/graders/07-anticipates-comeback.md 1 KB
- evals/work-email-push-back/graders/08-no-homework-for-user.md 1 KB
- evals/work-email-push-back/graders/09-no-personal-jabs.md 1 KB
- evals/x-condescending-expert/case.yaml 1 KB
- evals/x-condescending-expert/prompt.md 1 KB
- evals/x-condescending-expert/graders/01-however-pivot.md 1 KB
- evals/x-condescending-expert/graders/02-stacks-multiple-moves.md 1 KB
- evals/x-condescending-expert/graders/03-personal-jab.md 1 KB
- evals/x-condescending-expert/graders/04-short-enough-for-x.md 1 KB
- evals/x-condescending-expert/graders/05-facts-verified.md 1 KB
- evals/x-condescending-expert/graders/06-no-hard-line-violations.md 1 KB
- evals/x-condescending-expert/graders/07-anticipates-comeback.md 1 KB
- evals/x-condescending-expert/graders/08-no-homework-for-user.md 1 KB
- evals/x-condescending-expert/graders/09-dominant-stance.md 1 KB
- references/REFERENCE.md 8 KB
Versions
The database has no other skills named however.
Database snapshot: 13.11.2026 · package version 7cb01db8612b (SHA-256 digest of the entire package)
Downloads in the last 30 days: 0 · ranking
Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.
Similar skills
- 100slack-messaging
Slack channel and thread messaging skill with context-aware session management. Reads messages, sends replies, searches channels and users, drafts and schedules messages -- all via Slack MCP tools. Use this skill whenever the user wants to read Slack messages, reply in Slack,…