wild
Skill0 starsChanged 40 days agoMITView on GitHub
Use when the user says "go wild", "go crazy", "over the top", "no filter", "unhinged ideas", "wild ideas", or types /wild — they want maximum-range idea generation for a project, feature, name, plan or anything else, not a careful shortlist.
In this repository3 other files it gives its agents.
- brain-polishSkill
- cusSkill
- deskSkill
The file
Files
Wild
The user's other habits say argue first and pitch one practical thing. This skill is the opt-in exception: while it runs, don't argue, don't filter, don't hedge. Range is the whole point. The landing at the end is where judgment comes back.
The output (in this order)
1. The flood: 40+ ideas, one line each. Bold name + one sentence. No paragraphs; volume beats polish. Group them under these lenses, at least 4 per lens:
| Lens | Push it until… |
| Absurd scale | it serves 1 person forever, or 1 billion by Friday |
| Wrong medium | it's a physical object, a sound, a smell, a printed zine, a phone call, a game |
| Flip the user | it's for their parent, a first-year, a dog, their future self, an enemy, an AI |
| Collide projects | two of the user's projects fuse into one thing |
| Cog sci mode | it tests a real theory of mind, memory, attention or consciousness |
| Startup in 10 days | a 3-person team ships it and charges for it |
| Anti-idea | the opposite of what they asked for, done on purpose |
| Pure bit | it exists only because it's funny |
2. Mutate: pick the 5 strangest from the flood and push each one level further (bigger, weirder, more literal). One line each, labelled "→".
3. Land it: exactly two picks.
- Wild but doable this month: the one they could actually start this week, with the first concrete step.
- Pure fun: the one they should do only because it'd be a great story.
Then stop. No caveats section, no "of course, consider…".
Grounding
Pull specifics from the user's world so ideas are theirs, not generic: ~/Notes/INDEX.md, now.md, the relevant projects/*.md, ideas.md. Real names, real projects, real constraints (where they work, what they are studying, what they build with) make the weird ideas land.
Common mistakes
| Mistake | Fix |
| 10–15 sensible ideas, sorted by feasibility | 40+, sorted by lens; feasibility waits for the landing |
| Paragraph per idea | One line each; paragraphs kill volume |
| Every idea is software | The wrong-medium lens must include at least one physical thing |
| Skipping mutation, jumping to a pick | The best ideas usually come from mutating the strangest ones |
| Hedging ("this might be too much") | Delete it. He asked for too much |
| Ending with a pile of options | Exactly two picks, then stop |
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.