x-crisis
Skill1 starsChanged 5 days agoMITView on GitHub
Handle a post that went wrong on X - a pile-on, a ratio, a factual mistake, a misread joke, a screenshot of the user circulating out of context - and decide between correcting, clarifying, apologising, or saying nothing, then draft exactly that. Use when the user says "I'm getting ratioed", "this blew up badly", "people are angry at my post", "should I delete this", "how do I apologise", "someone quote-tweeted me to dunk", or is in the middle of a pile-on.
The file
We split this skill into the rules it gives an agent
8 rules. About 0 are common, 7 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈7
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 22
Ask, in one batched question, and do not draft anything until you have them:
- Distinctive: line 30
If there is legal or safety exposure, say plainly that this needs a person (a lawyer, an employer's comms team, the platform's reporting …
- Distinctive: line 46
Delete: when the post contains a factual error that could hurt someone, private information, or something the user now believes is wrong.
- Distinctive: line 49
Keep: when it is a defensible position people disagree with.
- Very specific: line 61
Every draft here is short and plain, then goes through /x-human and /x-preflight --kind reply.
- Distinctive: line 70
Never post, delete, block or report on the user's behalf.
- Distinctive: line 71
Never draft a reply that attacks a specific critic by name.
- Distinctive: line 72
Never invent context ("I was hacked", "it was a joke") the user did not confirm as true.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
x-crisis
The first hour of a bad post decides how long it lasts. Most of the damage in a pile-on comes from the second post: the defensive reply, the sarcastic quote, the deletion followed by a screenshot of the deleted post. This skill slows that down and picks one move.
First, the facts
Ask, in one batched question, and do not draft anything until you have them:
- The post, exactly as posted, and when.
- What people are saying, pasted: the ten most-liked replies and quotes, not the worst ones. The worst ones are not representative; the most-liked ones are what new readers see.
- Is any of it right? The user's honest answer, in private.
- What is at stake: employer, clients, legal exposure, someone's safety. If there is legal or safety exposure, say plainly that this needs a person (a lawyer, an employer's comms team, the platform's reporting tools), and stop here.
Then sort it
| what happened | the move |
| A factual error | Correct it, in a reply to the original, today. Say what was wrong and what is right. No "for those who misunderstood". |
| Right, but read wrong | One clarifying reply. If the same misreading appears in most top replies, the post was unclear, and that is on the post. |
| Wrong, and it hurt people | Apologise. One post, named harm, no "if anyone was offended", no explanation of intent. Then stop posting about it. |
| A disagreement, loudly | Nothing, or one reply to the best-faith critic. Volume is not evidence of being wrong. |
| Bad faith / brigading | Nothing. Mute, block, hide replies, report what breaks the rules. A reply gives it reach. |
Delete or keep
- Delete when the post contains a factual error that could hurt someone, private information, or something the user now believes is wrong. Say that it was deleted and why, in one line, because screenshots outlive deletions.
- Keep when it is a defensible position people disagree with. Deleting that reads as conceding the bad-faith version.
What the ranker does here
X's published weights put mute, block, "not interested" and report far below
zero (python3 ../x-plan/weights.py). A defensive reply or a dunk-quote
creates exactly those signals, from the people already annoyed. That is the
practical case for saying less.
Drafting
Every draft here is short and plain, then goes through /x-human and
/x-preflight --kind reply. A crisis reply with an em dash or a stock phrase
in it reads as a PR team, which is worse than silence.
Then a 48-hour plan: what not to post (anything about it, any subtweets), when normal posting resumes, and which replies to leave unanswered.
Never
- Never post, delete, block or report on the user's behalf.
- Never draft a reply that attacks a specific critic by name.
- Never invent context ("I was hacked", "it was a joke") the user did not confirm as true.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

