lenbok/qrspi

qrspi-question

QRSPI phase 1 — Decompose a task into neutral research questions

Quelltext ansehen

Verwendung

  1. Installationsbefehl kopieren und das Quellpaket vor der Installation in der eigenen Agentenumgebung prüfen.
  2. Benötigte Werkzeuge, Zugangsdaten und Hosts im Original prüfen.
  3. Eingabe und gewünschtes Ergebnis beschreiben und die Ausgabe mit den Quellen abgleichen.
npx skills add https://github.com/lenbok/qrspi --skill qrspi-question

Vor dem Start

  • Die Installation umfasst weder Modellabo noch API-Kontingent oder Dateizugriffsrechte.
  • Sterne und Installationen zeigen Aktivität, keine unabhängige Qualitäts- oder Sicherheitsprüfung.
Originales Skill-Dokument

Aus dem Quell-Repository gerendert; Überschriften, Beispiele, Code, Tabellen, Links und Bilder bleiben erhalten.

Invoke with: <ticket file, issue URL, or task description>

Question — Decompose the Task

Transform a task description into 3-7 specific, neutral research questions. These questions drive the next phase (Research) which runs in a separate context with no knowledge of what is being built.

Input

The user provides a task description, ticket file path, or issue reference.

Process

  1. Read any provided files fully before doing anything else.
  1. Light codebase exploration: Spawn a codebase-locator agent to find which areas of the codebase relate to the task. You need to know what exists to write good questions.
  1. Decompose into 3-7 research questions:
  • Each question should cause a researcher to explore a different relevant area of the codebase
  • Questions must be neutral — they ask what exists and how it works, never how to build something
  • Prefer "trace the flow" questions that reveal architecture over yes/no questions

Good: "How does the middleware chain handle request authentication, and where are auth policies defined?" Bad: "What's the best way to add a new authenticated endpoint?"

Good: "What patterns exist for database migrations, and how are they tested?" Bad: "How should we add a new migration for the users table?"

  1. Determine the artifact directory:
  • With ticket number: thoughts/qrspi/PROJ-1234-brief-description/ (use the project's ticket prefix)
  • Without ticket: thoughts/qrspi/YYYY-MM-DD-brief-description/
  1. Create the artifact directory (relative to the working directory) if it doesn't exist (e.g., mkdir -p thoughts/qrspi/<id>/).
  1. Write `task.md` — a clean 2-3 sentence description of what's being built and why. This file persists the task context for later phases so the user doesn't have to re-explain it.
  1. Write `questions.md` to the artifact directory:
markdown
   # Research Questions

   ## Context
   [2-3 sentences describing which areas of the codebase to focus on.
   Do NOT mention what is being built or why.]

   ## Questions
   1. [Neutral, fact-seeking question]
   2. [Neutral, fact-seeking question]
   ...
  1. Present questions to the user and wait for approval or edits before finalizing.

Output

  • Directory created: thoughts/qrspi/<id>/
  • Files written: thoughts/qrspi/<id>/task.md and thoughts/qrspi/<id>/questions.md
  • Tell the user: "Next: invoke the qrspi-research skill with thoughts/qrspi/<id>/"

Rules

  • questions.md must NOT contain the task description, goals, or desired behavior
  • task.md is a brief, honest description of the goal — it will be read by later phases but NOT by Research
  • The researcher who reads these questions should have no idea what feature is being built
  • Each question should target a different area or concern
  • If the task is too simple for 3 questions, tell the user — QRSPI is for complex tasks

Verwandte Aufgaben

aus demselben Repository

Weitere Skills

Alle Skills