Paperthin skill
28 skills that remove instead of add: rewrite a document clean, check the read of a request, compress text and challenge a decision
- 1,250 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Skills edit and overwrite your project files, for example re0 and ssotize
- re0-upgrade reinstalls skills and writes a session-start hook into agent settings
- re0-release and re0-merge deal with releases and merging someone else's code
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/paperthinDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Skills edit and overwrite your project files, for example re0 and ssotize
- re0-upgrade reinstalls skills and writes a session-start hook into agent settings
- re0-release and re0-merge deal with releases and merging someone else's code
This is third-party code. Review the repository files before installing.
What it does
Paperthin is a set of 28 skills with one idea: the agent removes surplus instead of adding files, options and boilerplate. The skills sit on a map with two axes, number of artifacts and time, in four folders: depth, breadth, coil and mesh. depth holds tools for a single artifact. re0 rewrites a drifted document into a clean version instead of another patch. readchk restates a request and asks only if a real fork remains. debloat compresses text to its load-bearing density, dedash removes em dashes, shower cold-reads material with fresh eyes and factchk checks claims against sources. hate finds the one objection that could kill a plan and the cheapest test for it. In breadth, ssotize folds a scattered fact into one home. The coil folder runs an iteration cycle: re0-plan, re0-loop, re0-memo, re0-work, catchup and nba. mesh holds prism, which splits one artifact across independent lenses and shows where they clash. The agent calls some skills on its own and others run only by name, for example /re0.
Who it is for. For people who run long projects with an agent and are tired of bloated documents, surplus edits and plans nobody stress-tested.
Good fit when
- A document has been edited many times and grown change history, and you want it rewritten as a clean first version
- A long multi-part request, and you want to be sure the agent read it right before it starts working
- A plan or decision looks convincing and you want its single strongest counterargument with a cheap test
- The same fact is scattered across several files and has to collapse to one source
- A release or a merge of someone else's contribution: re0-release and re0-merge walk a checklist with confirmation at each step
Not a fit when
- You need a ready tool for a specific stack such as a linter or code generator: Paperthin works on reasoning and text
- You want a step-by-step development method with TDD and subagents: other sets cover that
- You want the agent to add features rather than trim: the philosophy of the set is the opposite
- You are not ready for skills that edit files and change the local install
Example request
Run /re0 on docs/architecture.md: rewrite it as a clean first version with no trace of edits or duplicatesLimitations
The set relies on the agent's discipline: a skill guarantees nothing if the model ignores its instruction. Twelve skills, among them hate, macrothink, feynman, debloat, re0-release and re0-merge, fire only by a name a human types. The agent may start the others itself, so know about possible extra triggers in advance.
The re0-upgrade skill updates the whole install and writes a quiet session-start notice into each agent's config, so run it deliberately. The README and docs are in English, with translations including Russian under docs/readme.
How to disable. Remove the installed skills with the skills manager or by hand from the agent's skills directory. If you ran /re0-upgrade, also remove the session-start notice it wrote into the agent's settings.
FAQ
How do I install Paperthin in Claude Code?
Run npx skills@latest add LilMGenius/paperthin --global --agent '*'. The install makes symlinks, so skills update on their own; if the system asks for administrator rights, allow it. Call by name: /re0.
How is re0 different from ordinary editing?
re0 does not append over the old text; it rewrites the material as if it were the first version, removing edit traces, duplicates and stale details. If there is nothing to improve, it changes nothing.
Which agents are supported?
The README lists Claude Code, Codex, OpenCode, Antigravity, Copilot, Cursor, Grok-Build, Pi, Hermes and OpenClaw, and the --agent '*' flag installs the skills for all found.
How do I update the skills?
Call /re0-upgrade. It retires renamed skills, adds new ones and asks for confirmation before each change; it also sets up a quiet notice about new skills at session start.
What is the difference between model-invoked and user-invoked skills?
The agent can start model-invoked skills itself from the task description. User-invoked skills fire only when you type the name, which is done for actions such as committing, releasing and harsh plan critique.
Discussion
No questions or reviews yet.
Related
SkillsA skill plus hooks that keep a coding agent on task and cut speculative defenses, unread checksums and scope creep
A skill that distills a person's digital twin from chat logs and documents into a ready Agent Skill package
A multi-agent harness that verifies results by artifacts: stop hooks, an independent judge and an event log for Claude Code, Codex, Cursor and more
A completion-discipline skill for large agent tasks: write the acceptance ledger with runnable checks first, then do the work, then reverify against evidence
A skill based on the Recursive Language Models paper: the agent splits repo-wide or multi-PDF tasks into batches for sub-agents
A skill that cuts an agent's prose output, plus a local proxy that compresses what the agent reads
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Repository stars | Main difference |
|---|---|---|---|
| Paperthin (this entry) | no | 1,250 | |
| Karpathy-inspired Claude Code guidelines | no | 221,177 | four principles in a single CLAUDE.md, with no separately invoked skills |
| antislop | no | 6,514 | rules against the generic AI look in interfaces, text and code comments |
| Superpowers | no | 302,161 | sets a development process: plan, TDD, subagents, code review |
README in short
The README opens with a 15-second quick start, then shows the two-axis map and an index of all skills in four folders with columns for purpose, scope, invocation and a read-only flag. The problem section follows, where the author calls most skills slop, and then a list of fixes: each agent ailment has its own skill, prior art and proof.
README badge
Original SKILL.md
--- name: re0 description: "Refresh an existing artifact into the current best v0. Use when the user asks to clean up, sync up, dedupe, de-noise, smooth, rewrite, or update an artifact after iteration; when nearby artifacts may have drifted; or when changes in one place should be reflected across related artifacts while keeping the result minimal." --- Refresh the target artifact as if it were the first clean version. ## Goal The result must be lighter, more current, and more accurate than the input. It should not read like a changelog, cleanup note, or patch over an older draft. ## Workflow 1. Identify the target artifact from the user's request or the active context. 2. Read the target artifact end to end before editing. 3. Check nearby artifacts that must stay aligned with it. 4. Remove scaffolding residue, stale deltas, duplicated process noise, deprecated information, and over-specific history. 5. Fold durable lessons into the place they should have lived from the start. 6. Rewrite instead of appending when appending would preserve noise. 7. Preserve the artifact's useful voice and structure; simplify everything else. 8. Smooth prose noise: fold a parenthetical that interrupts a sentence or list into its own clause or cut it, keep a word or point repeated within reach of itself only once, and unwind padding that adds length but not meaning. 9. Smooth source noise: unwrap a hard line break that splits a sentence mid-flow so each paragraph or list item lives on one source line, matching how sibling artifacts format theirs; make a plain-text pointer to a named section or file followable, a link where the medium supports one; leave code, tables, and quoted material as formatted. 10. Re-read the result and cut again. ## Rules - A pass that finds nothing to genuinely improve changes nothing. - Prefer editing existing sections over adding new ones. - Convert "what changed" into "what is true now". - Keep only details that improve future execution, accuracy, or recall. - Do not leave old/new traces unless the artifact is explicitly a changelog. - Fix machine-clear residue; surface judgment calls — never auto-resolve an ambiguity or "fix" a deliberate look-alike. - Repair from the source of truth — the canonical or sibling artifact, not the damaged surface — and mutate with edit-safety: assert the target exists (report a MISS, never a silent no-op), edit unicode-safe (`PYTHONUTF8=1`), replace positional targets per occurrence rather than by blanket sweep, and script large structural moves. - Do not create extra files unless the user asks. ## Verification The result reads as a clean v0 to fresh eyes — no trace of the older draft, no sign it was patched rather than rewritten. If unsure, re-read it cold, as a stranger with no prior context would. Report what noise was removed and what durable truth was kept.