agentleFS
maroffomaroffo/claude-forge/skills/refine-requirements/SKILL.md

refine-requirements

Skill17 starsChanged 51 days agoMITView on GitHub

Structured requirements gathering before planning. Use when request has ambiguities, multiple valid approaches, or implicit decisions.

The file

We split this skill into the rules it gives an agent

2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Pick the scope mode: before any other question.

    line 15
  2. Distinctive:

    Follow threads: each answer may reveal the next question

    line 42

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

ABOUTME: Structured requirements refinement using AskUserQuestion before planning

ABOUTME: Identifies gray areas, clarifies with concrete options, captures decisions for the plan

Requirements Refinement

Invoke explicitly or triggered automatically by plan-first-workflow for 🟡/🔴 tasks.

Process

  1. Pick the scope mode before any other question. Classify the task, then:

    Task classModes offeredApplied
    bugfix, refactornone, no question askedHold Scope, silently
    featureHold Scope (recommended) / Selective Expansion / Reductionuser's answer
    greenfieldHold Scope (recommended) / Selective Expansion / Reduction / Expansionuser's answer
    unsure which classnone, no question askedHold Scope, silently

    One AskUserQuestion for the mode where modes are offered; Hold Scope is the recommended option. Never ask on a bugfix or refactor, and never ask when the class is unclear: Hold is the escape from misclassifying.

    • Hold Scope: clarify HOW to build what was requested, never WHETHER to add more (steps 1-4 below as written)
    • Selective Expansion: adjacent additions may be proposed, one at a time
    • Expansion: same, with a wider radius (greenfield only)
    • Reduction: propose cuts to what was requested, one at a time

    In Selective Expansion and Expansion: each proposed addition is its own AskUserQuestion; never bundle. Reduction works the same way for cuts: one question per cut, nothing dropped silently. Declined proposals become deferred ideas, exactly as in Hold Scope.

  2. Identify gray areas by domain:
    • Visual → layout, density, interactions, empty states
    • API/CLI → response format, error handling, auth flow
    • Infrastructure → scaling, redundancy, monitoring
    • Integration → protocol, auth method, error recovery
  3. For each gray area, use AskUserQuestion:
    • Concrete options ("JWT" not "Option A"), 2-4 choices
    • Include "You decide" when Claude's discretion is fine
    • Follow threads: each answer may reveal the next question
  4. Scope guard: in Hold Scope (and outside the opted-in mode's radius), if user suggests new scope, capture as deferred idea, redirect back
  5. Output: list of decisions to include as ## Decisions in the plan file

Anti-patterns

  • Generic questions ("What are your requirements?") → ask about specific decisions
  • Checklist walking → follow the thread the user cares about
  • Expanding scope in Hold Scope → clarify HOW to implement what's requested, not WHETHER to add more
  • Asking the mode question on a bugfix, a refactor, or an unclear task → Hold applies, silently
  • Bundling several additions into one "expansion accepted" → one question per addition, always

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.