agentleFS
shinprshinpr/codex-workflows/.agents/skills/requirement-convergence/SKILL.md

requirement-convergence

Skill39 starsChanged 10 days agoMITView on GitHub

Converges future-state requirements around outcome, exclusions, and rough cost. Use when users ask 'do we need this?', 'how far should we go?', or request a feature or scope change.

2 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

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

  • 4

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    All four fields apply whenever Requirement Convergence runs.

    line 25
  2. Distinctive:

    Use references/criteria.md to judge each field.

    line 27
  3. Very specific:

    Ask only about fields below ready.

    line 38
  4. Very specific:

    If an answer still fails its pass condition, ask once more.

    line 40
  5. Distinctive:

    Read the convergence record from the current handoff or its durable document.

    line 56
  6. Very specific:

    Keep weak-but-explicit fields visible as open questions.

    line 58

≈ “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

Requirement Convergence

Purpose

Capable models can make an oversized or misdirected request technically coherent and implement it faithfully. Converge what is worth building before design decides how to build it.

Apply this skill to future-state Design Doc and implementation flows. Reverse-engineered/as-is documentation is outside its scope. Requirement Convergence decides what to build; design then decides the smallest sufficient way to build the converged scope.

Convergence Record

FieldPass condition
outcomeOne observable result. Every buildable requirement serves it.
requirements[]Every item is labeled current-state, desired-future, or speculative; buildable outcomes and explicit constraints trace to the user's request. Agent-selected means remain design choices.
nonGoals[]The user decided each exclusion or explicitly stated there are none.
costA rough band with structural evidence and remaining unknowns.

cost is an early requirements estimate, not a design or work-plan estimate. It is intentionally approximate: enough to challenge low-value scope without pretending shallow inspection can produce exact effort.

All four fields apply whenever Requirement Convergence runs. Each field has readiness ready, weak, or weak-but-explicit. Only the user can accept weak-but-explicit. The record is converged when every field is ready or weak-but-explicit.

Use references/criteria.md to judge each field.

Hearing Protocol

The orchestrator owns user interaction. It runs the hearing after an analysis step has produced scope facts.

The user decides product requirements and exclusions. Keep user-selected scope, observed facts, and agent inference separate because combining them makes agent interpretation indistinguishable from evidence and prevents an informed requirement decision.

At every requirements confirmation stop, render references/scope-confirmation.md. Only an explicit user answer moves a user-owned decision into user-selected scope.

  1. Escalate a requirement change only when concrete evidence shows the requested outcome cannot be achieved under the current conditions; name the failing premise, smallest change, and trade-off for the user's decision.
  2. Ask only about fields below ready.
  3. Record answers in the user's wording.
  4. If an answer still fails its pass condition, ask once more. Mark it weak-but-explicit only when the user agrees to proceed unresolved.
  5. Re-judge the updated record before design begins. Re-run structural analysis only when an answer changes scope or cost evidence; otherwise the orchestrator applies the field's pass condition directly.

Storage Protocol

Flow stateCarrier
Before persistenceCompact convergence object in the current handoff
PRD flowSuccess Criteria holds outcome; the requirement boundary holds current requirements and user-decided nonGoals
Design Doc is the first durable documentOverview holds outcome; Requirement Boundary holds current requirements and user-decided nonGoals
Small direct implementationCompact record embedded in the single task file's Governing Sources

Persist weak-but-explicit outcome, current requirements, and non-goals as open questions. Keep speculative ideas, evaluation requests, prescribed mechanisms, and unselected candidates in the active convergence context only until the user selects scope; durable downstream documents contain the resulting current requirements and selected conclusions. cost is also ephemeral: use it for the requirements challenge, then let Structural Scale select the workflow and let design or planning produce later estimates. A Small-flow task file carries only outcome, current requirements, nonGoals, and their readiness in Governing Sources; its executor receives only that task path. Keep relevant user wording and explicit constraints in that existing requirement boundary, distinct from technical means. Downstream agents can then use the document path; before persistence or when provenance is missing, pass the relevant source excerpt.

Downstream Contract

  1. Read the convergence record from the current handoff or its durable document.
  2. Build the current change from user-selected desired-future outcomes and keep recorded nonGoals outside implementation. Technical means remain revisable after phase passage when user-required outcomes and explicit constraints remain satisfied; reconcile affected sources and proof. A speculative capability becomes buildable only when the user explicitly requests that scope change.
  3. Keep weak-but-explicit fields visible as open questions. Escalate only when the current work depends on resolving one.

Quality Checklist

  • Scope facts and inferences are distinguished
  • Every buildable requirement traces to the outcome
  • Every non-goal is user-decided, and an empty nonGoals list reflects the user's explicit choice of none
  • Cost is approximate and evidence-backed
  • Every field is ready or user-approved weak-but-explicit

References

  • references/criteria.md — field judgments, cost bands, challenge intensity, and solution-in-disguise test
  • references/scope-confirmation.md — requirements-stop output shape

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.