lilmgenius/paperthin

re0-loop

Run repeated cycles of build, QA, re0-memo and re0-work while preserving learning and letting code die.

ソースを見る

使い方

  1. インストールコマンドをコピーし、自分の環境へ導入する前にソースを確認します。
  2. 必要なツール、認証情報、対応環境を原文で確認します。
  3. 入力と成果物を指定し、出力を資料と照合します。
npx skills add https://github.com/lilmgenius/paperthin --skill re0-loop

開始前の確認

  • 導入にはモデル契約、API枠、ファイルへのアクセス許可は含まれません。
  • スター数や導入数は活動の指標であり、独立した品質・安全性試験ではありません。
リポジトリの原文

見出し、例、コード、表、リンク、参照画像を含む原文を表示しています。

Run a long project through repeated laps of the coil cycle so learning compounds and code is free to die.

Goal

Keep a long project moving through repeated turns so learning compounds and piling up code never passes for progress. Use it on long agentic projects where progress must be measured by quality-cleared output rather than time or feature count. Progress counts only as quality-cleared templates, reusable platform modules, and anti-patterns eliminated in later cycles; never hours, files, panels, or features.

Each lap runs this cycle:

text
FRAME -> BUILD -> DRIVE -> RE0-MEMO -> HATE -> RE0-WORK -> BUILD AGAIN

Workflow

Run the steps in this order.

  1. Frame. Write the thesis and the quality gates the lap will be judged against.
  2. Build. Build one complete vertical slice.
  3. Drive. Drive the slice through its real surface, picked by artifact: a browser for web apps, HTTP for API contracts, desktop UI automation for desktop apps, and the CLI only for data-shaped artifacts. Capture what the drive shows as the lap's evidence.
  4. Memo. Call the Skill tool with "re0-memo" to extract lessons, anti-patterns, and next-cycle gates from the lap. The memo lands where that skill's own contract puts it.
  5. Decide. You decide whether the next pass iterates in place or restarts, and you state the choice as keep, iterate, or re0-work. For a restart, call the Skill tool with "re0-work". When that call, or the next move, is unclear, call the Skill tool with "nba", which reads cycle state and returns the single next action.
  6. Kill the plan. Attack the next plan before building it. Use the project's adversarial review skill. When that skill is user-invoked only, as hate is, the attack is the human's to fire: tell the human to run it, and never call it yourself.
  7. Version. Put a version label only on the templates or modules that cleared their gates. The next lap builds again from what survived.

Rules

  • Tests support; they are not the gate. Every lap needs a real-surface proof sized to the artifact.
  • Do not widen scope before the core loop clears.
  • Generated variety is not structural variety: if many outputs converge on one shape, the cycle failed the variety gate.
  • Preserve negative corpus so later cycles can prove an anti-pattern is gone.
  • Let code die: a restart that keeps the right lessons counts as progress.
  • A request the human makes while a lap runs goes onto the live plan the moment it arrives, beside earlier ones rather than over them, and leaves it only when it is done or the human drops it. A summary written when the context is compacted or handed to a new session lists every request still open, started or not, because a request held only in the conversation is lost with it.
  • Outside truth is evidence the lap did not produce about itself: a real-surface drive, a user's reaction, data from outside the build. A lap with none stops; add such evidence before running another pass rather than looping on itself.
  • Through the Skill tool, call only the model-invoked "re0-memo", "re0-work", and "nba". Route the plan attack to the project's adversarial skill, or to the human when that skill is user-invoked; never call a user-invoked skill.

Verification

Before closing each lap:

  1. The real surface was driven and evidence captured.
  2. The re0-memo names at least one lesson, anti-pattern, or gate that affects the next pass.
  3. The keep / iterate / re0-work decision is explicit.
  4. Any version label sits only on a quality-cleared template or module.
同じリポジトリから

関連する Skills

すべての Skills
lilmgenius
コミュニティ

re0-order

Move items only, reword nothing; realign a listing whose order has gone arbitrary (a list, table, catalog, sections, an enum) into a logical sequence under one stated principle. Nothing is added or removed. Use when the user names a listing, or a scope holding one, whose order no longer helps a reader follow the set, or when a pipeline the user invoked passes that scope on. Acts only inside that scope.

1時間の導入数
5
GitHub Stars
1463
更新日
11月15日
lilmgenius
コミュニティ

readback

Verify the model's understanding of a user's instruction before spending non-trivial work. Use when a request is long, bundled, high-stakes, hard to undo, or has ambiguous scope or referents such as this, that, it, the other one, whatever is cleaner, or whichever order makes sense. Restate internally, cross-check against available context, proceed silently when resolved, and surface only a genuine surviving fork.

1時間の導入数
5
GitHub Stars
1463
更新日
11月15日
lilmgenius
コミュニティ

whychk

Right after you pick an option, have a fresh outsider find the sharpest gap in the choice, then explain that gap in your own words, one gap at a time.

1時間の導入数
5
GitHub Stars
1463
更新日
11月15日
lilmgenius
コミュニティ

aim

When data lands with a thin ask (files, a repo, a paste, a chat scrollback), read it and propose what the user is trying to do with it for them to confirm. Use the moment material arrives without a fully-formed ask, or when the ask reads thinner than the data suggests.

1時間の導入数
4
GitHub Stars
1463
更新日
11月15日