菜单
Transform polished AI-generated text into authentic human-sounding copy. Use when user wants to "de-slop," "humanize," "de-AI," "make sound human," "add personality to," or "degunk" any AI-written content—emails, LinkedIn posts, marketing copy, etc. Also triggers on requests to make text sound "less robotic," "more natural," or "like a real person wrote it."
Post-generation diversity audit that scores whether a set of ideas are actually different from each other or just surface variations of the same thought. Use AFTER any brainstorming, ideation, or creative generation — from any source, any skill, any person — when you need to know if your "ten different ideas" are actually ten ideas or three ideas wearing costumes. Triggers on "are these actually different," "these all feel the same," "did I just write the same idea five times," "check these for diversity," "which of these overlap," or any moment after generating a list where the user wants to know what they actually have before moving forward. Also use automatically as a quality gate after any other creative skill produces output, or when a user pastes in a list of ideas from anywhere and wants them audited. Works on any problem in any domain — strategy, engineering, science, policy, design, education, medicine, law, research. This is the bullshit detector for brainstorms.
Generate deliberately terrible ideas, then reverse-engineer the hidden value inside them. Use as a creative de-inhibitor when the user or the AI is stuck producing safe, polished, "good" ideas that all feel the same. Triggers on "I'm stuck," "everything I come up with is boring," "I can't get past the obvious," "give me something wild," "help me loosen up," or any moment where the quality filter is killing creativity before it starts. Also use when a brainstorm feels constrained, when every idea sounds like it came from a committee, or when the user needs permission to be ridiculous before they can be brilliant. Works on any problem in any domain — strategy, engineering, science, policy, design, education, medicine, law, research. This skill is not about being random. It's about using intentional badness as a door to territory that "good" thinking can never reach. Generating bad ideas on purpose is psychologically different from generating good ideas and hoping some are weird.
Map the full problem space before or after ideation to identify what you forgot to think about. Use BEFORE generation to ensure brainstorming covers the whole territory, or AFTER generation to check whether your ideas left entire regions of the problem unaddressed. Triggers on "what am I missing," "what haven't I thought of," "is there a blind spot here," "what else should I consider," "did I cover everything," or any moment where the user suspects their thinking has been thorough within a narrow band but oblivious to adjacent territory. Also use when a brainstorm feels solid but suspiciously focused, when an Anti-Homogeneity Check returns a high grade but the ideas still feel incomplete, or when a strategy covers the offensive plan but has no defensive layer. Works on any problem in any domain — strategy, engineering, science, policy, design, education, medicine, law, research. This is the skill that asks not "are your ideas different?" but "what parts of the problem did you forget existed?"
Generate questions, not answers. Use when the user needs to reframe a problem, explore a space they don't fully understand, challenge assumptions in an industry or process, or find the door that better ideas could walk through. Triggers on "what questions should I be asking," "what am I not seeing," "help me think about this differently," "what would a beginner ask," or any moment where the user is stuck not because they lack solutions but because they're asking the wrong questions. Also use when a problem space feels stale, when experts have been working on something for years without progress, when an industry or field does things "because that's how it's done," or when the user says "I don't even know what to ask." Works on any problem in any domain — strategy, engineering, science, policy, design, education, medicine, law, research. This skill produces questions that open doors. Other skills walk through them.
Force genuinely divergent ideation using Guilford's five-dimension framework, backed by empirical research from Nature Communications (LiveIdeaBench). Use when generating ideas, exploring approaches, developing hypotheses, finding angles, designing solutions, or any task where the user needs multiple genuinely different options — not surface variations of the same thought. Triggers on "brainstorm," "give me ideas," "what are my options," "explore possibilities," "think broadly," "what else could this be," "I need approaches," or any request for a list of options. Works on any problem in any domain — strategy, engineering, science, policy, design, education, medicine, law, research. Use INSTEAD of generic list-making. The difference between this skill and a normal brainstorm is the difference between asking five people in the same room and asking five people in different buildings.
Force creative solutions using only the resources already available — no new budget, no new hires, no new tools, no new anything. Use when the user is stuck because they think they need something they don't have, when budget is zero, when time is short, when "we can't afford to" is the refrain, or when the obvious next step is "get more resources" and that's not an option. Also triggers on "I have no budget," "we're working with what we've got," "what can I do with what I have," "we can't hire anyone," "bootstrap this," or any moment where constraints feel like the problem but might actually be the solution. Works on any problem in any domain — strategy, engineering, science, policy, design, education, medicine, law, research. The best ideas in history were built from leftovers. This skill forces you to inventory what's already in the room and build from that — nothing else.
| name | de-slop |
| description | Transform polished AI-generated text into authentic human-sounding copy. Use when user wants to "de-slop," "humanize," "de-AI," "make sound human," "add personality to," or "degunk" any AI-written content—emails, LinkedIn posts, marketing copy, etc. Also triggers on requests to make text sound "less robotic," "more natural," or "like a real person wrote it." |
Transform sterile AI output into text that reads like it was written by a real person—someone with deadlines, coffee stains, and a working knowledge of when to break grammar rules.
AI text fails the human test not because it's wrong but because it's too right. Real humans:
Ask for intensity level and context. Present the options, then pause with a bold question so the user knows it's their turn:
Which level, and where is this going? (Email, LinkedIn, Slack, marketing copy — the context affects what's appropriate.)
Do not proceed until the user answers. Do not default to Medium. The choice is theirs.
Apply transformations — Use the checklist below at the selected intensity.
Show before/after — Present the original and transformed version for comparison.
Apply these in order, calibrated to the selected intensity level.
Kill on sight:
Banned openers:
Banned structures:
Glib colon phrases (kill all):
Dramatic single-sentence paragraphs: AI loves these. For emphasis. Over and over. Stop it.
Also kill:
Banned openers:
Break parallelism (occasionally):
Add sentence fragments. For emphasis. Like this.
Start sentences with "And" or "But" — real writers do this constantly.
Kill the wrap-up bow:
Interrupt yourself:
Reduce em-dashes — AI overuses these — like crazy — stop it
Inconsistent Oxford comma:
Imperfect ellipses:
Spacing artifacts:
Grammar "mistakes" that pass spellcheck:
Common typos:
Abbreviation chaos:
Filler injection:
Softeners:
Casual transitions:
Real writing has varying levels of effort:
Light (professional but warm):
Medium (casual professional):
Heavy (stressed middle manager at 11pm):
AI Original: "I wanted to reach out regarding the Q3 marketing strategy. After diving into the data, it's worth noting that our comprehensive approach has yielded robust results. Moving forward, I believe we should leverage these insights to navigate the evolving landscape. Let me know your thoughts."
Humanized (Medium): "Hey—wanted to loop you in on Q3 marketing stuff. Looked at the numbers and honestly the results are solid. Think we should use what we learned here as things keep shifting. lmk what you think"
Humanized (Heavy): "hey quick note on Q3 marketing. looked at the data and its actually pretty solid?? think we shoudl use this going forward, the whole space is shifting anyway so. lmk ur thoughts when you get a sec"
