re0-import
Land handed-over material of any shape (a note, a link, a dump, a correction) in the document of an existing knowledge base that…
Check whether an item falls inside a target's existing scope, as the target's own written sources define it; it never sets or narrows the scope of a piece of work. Use before a skill, script, module, rule, feature, doc section, or file is added to, moved into, promoted into,
Claude Code
$ npx -y skills add LilMGenius/paperthin --skill scopechk --agent claude-code
Codex
$ npx skills add LilMGenius/paperthin --skill scopechk -a codex -y
Cursor
$ npx skills add LilMGenius/paperthin --skill scopechk -a cursor -y
Gemini CLI
$ npx skills add LilMGenius/paperthin --skill scopechk -a gemini-cli -y
GitHub Copilot
$ npx skills add LilMGenius/paperthin --skill scopechk -a github-copilot -y
OpenCode
$ npx skills add LilMGenius/paperthin --skill scopechk -a opencode -y
Windsurf
$ npx skills add LilMGenius/paperthin --skill scopechk -a windsurf -y
name: scopechk description: "Check whether an item falls inside a target's existing scope, as the target's own written sources define it; it never sets or narrows the scope of a piece of work. Use before a skill, script, module, rule, feature, doc section, or file is added to, moved into, promoted into, merged into, or upstreamed to a suite, repository, plugin, package, or document, and when someone asks whether X is in scope for Y, whether X belongs in Y, or which of several homes should own X."
Check whether an item falls inside a target's existing scope before it lands there.
An item tends to land in whichever home is open, and once there it reads as part of that home's contract. `scopechk` returns a placement verdict for one item against each candidate target, read from what the target already says about itself: its written sources, or what its current members share where those are silent. It checks an item against a scope that exists and never sets or narrows the scope of a piece of work. It proposes; a human decides and moves.
1. **Pin.** Name the item, each candidate target, and the action asked (add, move, promote, merge, upstream). Separate any pointer the item carries (a link, a one-line reference) from the thing it points at; each gets its own verdict. If the item or a target cannot be read, report a MISS for it and stop there. 2. **Read the scope.** From the target's own sources (README, AGENTS.md, CONTRIBUTING, its layout cut, naming rules, install surface, stated audience), record each criterion with its file and section. Where the sources are silent, infer the criterion from what the current members share, label it `inferred`, and name those members. A criterion only the target's owner could settle becomes an owner question. 3. **Compare the nearest member.** Pick the existing member that serves the same kind of user or does the same kind of work, and say why it is nearest. Compare the item with it on each criterion. If no member is comparable, say so; that absence is itself evidence against the fit. 4. **List what the target takes on.** Write an inventory, not a score: new dependencies, the item reaching everyone who installs or clones the target, each public promise it would have to keep (a stable install, a release cadence, a public surface), and the upkeep it adds. 5. **Check reachability.** List everything the item points at (repositories, paths, services, accounts, tools) and confirm the target's users can reach each one, by opening it as they would or reading its visibility. Anything they cannot reach, such as a private repository in a public target or a path on one machine, is a failure. Anything you cannot check is marked `unverified` and named. 6. **Give the verdict,** one per item, pointer, and target:
7. **Stop.** End with `Approval stop`: the proposed placement of each part, any owner questions, and a statement that nothing was moved, copied, renamed, registered, or deleted. The human who asked decides whether and where it lands.
1. Every criterion cites its location in the target or carries the `inferred` label with its source members. 2. The nearest member is named with the reason it is nearest, or its absence is stated. 3. Every pointer was checked for reachability or marked `unverified`, and no failure sits under `in scope`. 4. Every verdict other than `in scope` names its change, its homes, or its owner question. 5. The run ends at the approval stop with nothing moved.
The skillbook for recursive self-improvement (RSI).NVIDIA's latest proof, unpacked with timeless engineering.Distilled from the OpenAI harness engineering hackathon winner's zero-touch method.
Land handed-over material of any shape (a note, a link, a dump, a correction) in the document of an existing knowledge base that…
Keep a repository's README for the person deciding whether and how to use it, its AGENTS.md (or a CLAUDE.md read through it) for…
Keep one paperthin install current and loaded once: move an old skills.sh install onto the Claude Code or Codex plugin, clear the…
Consolidate duplicate files or directories into one canonical home after approval, held as an identity by this machine's…
Audit and consolidate a fact that's scattered across places into one canonical source, after approval, and replace the rest with…
Feed what a finished cycle taught back through the sources that produced it: measure the lesson, question the requirements it…
Behavioral guidelines to reduce common LLM coding mistakes. Use when writing, reviewing, or refactoring code to avoid…
Design and build AI agents for any domain. Use when users: (1) ask to "create an agent", "build an assistant", or "design an AI…