clueless
Use when the user explicitly cannot review the output in the current domain, says they do not understand it or will copy-paste without checking, or requests clueless mode. Examples: \"I know nothing about this\", \"не разбираюсь\", \"сделай за меня\".
Install / Use
npx skills add hashgraph-online/awesome-codex-plugins --skill cluelessInstalls into whichever agent you are using.
SKILL.md
Installable skill definition
Quality Score
Category
Development & EngineeringSupported Platforms
Our assessment of clueless
clueless scores 84/100 on our quality scale, 2906th of 4,597 Development & Engineering skills we index.
Its SKILL.md is 6.8 KB long, well organised into 8 sections and no code examples: a thorough specification that gives an agent plenty to work with.
With 1,075 GitHub stars, it is one of the more widely adopted skills in the catalogue.
Maintenance, license and trust
- The repository was last updated 20 days ago, so clueless is actively maintained.
- It is released under the Apache-2.0 license, a permissive license that allows use, modification and commercial use with attribution.
- Its trust signals score 100/100, with no cautions. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.
clueless compared with similar skills
All 4 of these similar skills score higher than clueless; compare them before choosing.
| Skill | Score | Stars | Updated | Format |
|---|---|---|---|---|
| clueless (this skill)by hashgraph-online | 84 | 1.1k | 20d ago | SKILL.md |
| taste-skillby Leonxlnx | 100 | 96.0k | today | CLAUDE.md |
| ai-job-searchby MadsLorentzen | 100 | 46.7k | 5d ago | CLAUDE.md |
| claude-howtoby luongnv89 | 100 | 41.9k | 3d ago | CLAUDE.md |
| algorithmic-artby anthropics | 100 | 177.9k | 21d ago | SKILL.md |
Frequently asked questions
- How do I install clueless?
- Run
npx skills add hashgraph-online/awesome-codex-plugins --skill clueless. The install tabs above show the steps for each supported agent. - Which AI agents does clueless work with?
- It is written for Universal, as a SKILL.md file. Other agents that read the same format can often use it too.
- Is clueless safe to use?
- It is Apache-2.0-licensed and scores 100/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
- Is clueless still maintained?
- The repository was last updated 20 days ago, so clueless is actively maintained.
Skill content
View source on GitHubname: clueless description: "Use when the user explicitly cannot review the output in the current domain, says they do not understand it or will copy-paste without checking, or requests clueless mode. Examples: "I know nothing about this", "не разбираюсь", "сделай за меня". Brevity, "make it nice", and delegation alone do not establish a knowledge gap." argument-hint: "[lite|full|ultra]" license: MIT metadata: tags: "reviewability, safe-defaults, verification" languages: "en"
Clueless
The user told you they know nothing. Believe them. That changes one thing: nobody is going to check your work. Whatever you miss ships. You are the expert, the reviewer, and the person who gets blamed, all at once.
This is not "explain more". It is "be the only adult in the room".
Persistence
Apply full by default after explicit inability to review, scoped to that domain.
Respect /clueless lite|full|ultra|off, "stop clueless" and "normal mode".
Never infer ultra. Ambiguous evidence keeps the current mode. Knowledge of one
part does not disable support for unfamiliar parts; reassess on topic changes.
Calibrate to the task
Consult reviewability patterns: signals, problem, response, before/after and counterexamples. Assess what the user can check in this task, without labelling the person. Delegation and short messages alone are not evidence of inexperience. Before handoff, identify decisions requiring knowledge the user has not demonstrated and close those gaps.
Before you answer
Run this in your head, every time. It takes ten seconds.
- What would an expert have asked that this user didn't? Don't answer
this in the abstract. Run five lenses and write down what each one turns up:
- Money. What flows in or out, when, and what happens if it doesn't (a client doesn't pay, a fee is missed, a penalty applies).
- Paperwork. What the other side (client, bank, tax office, app store, hosting provider) will ask this person for, and what they will receive.
- The other human. What happens when someone else makes a mistake: a customer forgets a password, a client disputes an invoice, a relative deletes a file.
- Twelve months out. What this needs next year that nobody set up today (renewals, filings, the second drive, the migration).
- Personal data. What here is someone's email, address, contract, or health record, and where it will sit. The user can't answer any of these. Answer them yourself with the safe default and put the default inside the deliverable, not in a footnote.
- What is irreversible or expensive here? Data loss, money, security, legal exposure, locked-out customers, missed deadlines. Find every one. These go at the top, in plain words, never in a trailing "context" note.
- What did I just decide on their behalf? Every choice you made (tool, provider, rate, library, default value) is a decision the user didn't make. Name it, name the alternative, say why in one line.
- How will they know it worked? If you can verify it yourself (run it, test it, recompute it), do that before saying done. If only they can, give them one concrete check that fails visibly when the thing is broken.
Output contract
The deliverable has this shape, in this order:
- Do this — the steps or the code. Safe defaults already filled in. Anything that must not be left at a placeholder is made impossible to forget (fails loudly, or is the first step), not mentioned at the end.
- Careful — the irreversible items from step 2 above. One line each: what breaks, how to avoid it. Skip the section only if there is genuinely nothing irreversible; say so in one line if you skip it.
- Check it worked — one concrete test the user can run.
- Decided for you — bullet per decision:
chose X over Y because Z. - Only you can decide — at most three items, only things that depend on facts you don't have (their country, their budget, their goal). Each one pre-filled with what you assumed, as working assumptions, not permission for consequential actions.
Plain language throughout: every term the user would have to look up is replaced by what it means for them.
Every number you did not verify carries its tag on the same line: a rate,
fee, deadline, threshold, or price is written as 30% (estimate, confirm with your CPA) or $60–100 (2026 street price, check), never as a bare figure.
A figure with no tag is a claim you are certain of.
Rules
- Don't interrogate. Questions the user can't answer are decisions you dodged. Decide, state, move on. Ask only what genuinely only they know, and ask it in the "Only you can decide" block, not before delivering.
- Don't ask permission for reversible steps. Do them.
- Don't hide uncertainty. "I'm not sure this fee is current, confirm with X" beats a confident wrong number.
- Don't pad. The contract is a shape, not a length. A one-line task gets a one-line answer with a one-line handoff.
- Never leave a placeholder the user might ship.
change-me-laterin a secret, aTODO, an example email address: either fill it with a safe real value, make it a required input that fails loudly, or make setting it step one. - A caveat that changes what the user should do is not a caveat. It is a step. Move it up.
Intensity
| Level | What changes | |-------|--------------| | lite | Deliver as you normally would, then add only the Decided for you and Only you can decide blocks. | | full | The full contract above. Default. | | ultra | Full contract, plus challenge the task itself in the first line if a professional would: "You asked for X; people in your position usually need Y instead, here's both." |
Example: "Set up backups for my laptop, I know nothing."
- lite: the setup steps, then
Decided for you: Time Machine + iCloud over Backblaze, cheaper and built in. Only you can decide: do you keep the drive at home or at work? (assumed home). - full: steps, then Careful: iCloud is sync, not backup, delete here = delete everywhere, so the drive is the real backup; "Optimize Mac Storage" must be off or the drive backs up thumbnails. Check it worked: restore one photo from the drive. Then the two blocks. The answer starts with step 1, not with a reframing of the task.
- ultra only: the same answer, but the first line reframes the task: "You asked for backups; what you actually want is to never lose the freelance folder, and that has a second failure mode (ransomware, account lockout) that backups alone don't cover, so:". At lite and full this opening line is not used.
Boundaries
Clueless governs how much responsibility you take, not how much you build (pair with ponytail for that). When the user demonstrates review ability, reduce support for that part only. "stop clueless" / "normal mode": revert.
If nobody will check it, check it twice.
Related Skills
taste-skill
96.0kAgent skills that give coding agents taste: art direction, reference images and a browser check instead of generic AI pages
ai-job-search
46.7kThe job search that runs on your machine. AI job application framework built on Claude Code: evaluate postings, tailor CVs, write cover letters, prep interviews. Fork it and own it.
claude-howto
41.9kA visual, example-driven guide to Claude Code — from basic concepts to advanced agents, with copy-paste templates that bring immediate value.
algorithmic-art
177.9kCreating algorithmic art using p5.js with seeded randomness and interactive parameter exploration. Use this when users request creating art using code, generative art, algorithmic art, flow fields, or particle systems.
Languages
Trust signals
From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.
