agentleFS
shinprshinpr/ai-coding-project-boilerplate/.claude/skills-en/requirement-convergence/SKILL.md

requirement-convergence

Skill239 starsChanged 3 days agoMITView on GitHub

Separates the outcome a change must produce from the requirements proposed to reach it, records what the user excluded, and bands cost from structure. Use when a requirement enters a workflow, before design begins, or when "how far do we go/what's out of scope/is this worth it" is mentioned.

2 other projects use a version of this, 1 of them edited.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 2 are distinctive, and 2 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.

  • 2

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    When neither document exists, retain the record in the active context.

    line 53
  2. Distinctive:

    Read the convergence record from the prompt.

    line 57
  3. Very specific:

    Treat nonGoals as excluded from the current change and desired-future requirements as buildable scope.

    line 58
  4. Very specific:

    Treat a weak-but-explicit field as a recorded open question rather than a settled decision, and escalate when the work depends on resolving …

    line 59

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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

Requirement Convergence

Purpose

Requirements arrive bloated, ambiguous, or aimed at the wrong outcome. A capable model reconciles all three into a coherent plan and builds it faithfully — delivering exactly what was asked for when what was asked for was wrong.

This skill converges what to build. How to build it, and which documents the change requires, come after the what.

Convergence Fields

FieldPass condition
outcomeOne observable result. A requirement that does not serve it is excess.
requirements[]Every build-relevant item labeled current-state or desired-future.
nonGoals[]Authored by the user, or the user stated there are none.
costA band with the structural evidence that places it, plus the unknowns that remain.

cost is a rough band, not the effort estimate a work plan schedules against; requirements cannot support person-days. Its unknowns carry more decision weight than its size.

Classify from the user's own retained wording rather than an analyzer's restatement of it. Wording that asks for an evaluation, describes a speculative idea, or suggests a mechanism stays in active convergence context as a judgment-only candidate. requirements[] and durable documents receive a candidate only after explicit user confirmation.

Each field carries its own readiness label: ready, weak, or weak-but-explicit (weak, and the user agreed to leave it unresolved). Only the user sets weak-but-explicit. Requirements are converged when every applicable field is ready or weak-but-explicit.

Judgment rules per field: references/criteria.md.

Hearing Protocol

Use the available scope and cost evidence for both elicitation and judgment, then ask only for product choices that the repository cannot answer. Build requirements[] and the exclusion hearing from capabilities the user requested; an adjacent capability noticed during analysis stays out of both, so the user need not reject a capability they did not request. Repeat scope and cost analysis only when an answer changes the analysis target or required scope evidence.

Register these steps before starting and record each step's evidence as it completes:

StepActionCompletion evidence
1State the available scope facts, then separately what they imply for the requirementFacts listed with their analysis evidence
2Ask about the fields below ready, at most two questions per messageOne question per field below ready
3Record each answer as that field's valueThe value is the option the user explicitly selected or the wording the user supplied
4Re-ask once when a recorded value still fails its pass condition, then mark the field weak-but-explicit when the user agrees to leave the second answer as it standsTwo recorded answers, or the user's agreement to stop
5Judge each field against its pass condition and finalize the completed recordA convergence record with every field labeled

Storage Protocol

CarrierHolds
The active convergence recordEvery field with its readiness label
PRD Success Criteria and Out of Scopeoutcome; user-authored nonGoals
Design Doc Requirement ConvergenceThe same when no PRD exists, and the fields left weak-but-explicit in every case

When neither document exists, retain the record in the active context.

Reference Protocol

  1. Read the convergence record from the prompt.
  2. Treat nonGoals as excluded from the current change and desired-future requirements as buildable scope. Evaluation requests, speculative ideas, prescribed mechanisms, and agent-proposed capabilities that were not promoted create no implementation obligation; an accepted ADR may retain evaluated options as decision history.
  3. Treat a weak-but-explicit field as a recorded open question rather than a settled decision, and escalate when the work depends on resolving it.

Quality Checklist

  • Scope facts were presented before questions were asked
  • nonGoals came from the user, or the user stated there are none
  • Every applicable field is ready, or weak-but-explicit by the user's agreement

References

  • references/criteria.md — judgment rules per field, cost inputs, challenge intensity, solution-in-disguise test

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.