requirement-convergence — Claude Code skill: install and overview
requirement-convergence — a Claude Code skill that isolates a project outcome from proposed requirements and records user-stated exclusions. Runs at requirement intake before any system design begins.
MIT ★ 720 updated 2026-11-11
npx skills add shinpr/claude-code-workflows --skill requirement-convergence -a claude-codeWhat requirement-convergence does
The skill standardizes what to build before teams decide how to build it. It breaks down input into an observable outcome, build-relevant requirements[] marked as current or desired future state, user-defined nonGoals[], and a rough cost band based on repository evidence.
During scope analysis, the hearing protocol questions only fields rated below ready. Answers are captured strictly in the user's wording rather than model summaries. If a field fails its pass condition twice, the user can mark it weak-but-explicit to leave it deliberately open.
The resulting convergence record writes to PRD sections, design documents, or the workflow context. Speculative agent suggestions and unconfirmed mechanisms are excluded from downstream execution tasks.
What it helps with
- separate business outcomes from bloated or premature implementation details
- record user exclusions as non-goals in the user's exact phrasing
- band implementation cost and lingering unknowns from repository evidence
- block agent-proposed mechanisms from becoming unintended build scope
Example prompts for the agent
- Run requirement convergence on the new notification service before writing the design doc
- Extract the outcome, requirements, and nonGoals for the billing migration task
- Check the readiness of our scope fields and present questions for items below ready
How to install requirement-convergence
1Skills CLI (Vercel)
npx skills add shinpr/claude-code-workflows --skill requirement-convergence -a claude-code2Manually: put the files in place
~/.claude/skills/requirement-convergence/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/requirement-convergence/
1Skills CLI (Vercel)
npx skills add shinpr/claude-code-workflows --skill requirement-convergence -a codex2Manually: put the files in place
.agents/skills/requirement-convergence/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add shinpr/claude-code-workflows --skill requirement-convergence -a cursor2Manually: put the files in place
.cursor/skills/requirement-convergence/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install shinpr/claude-code-workflows dev-skills/skills/requirement-convergence2Manually: put the files in place
.github/skills/requirement-convergence/SKILL.md1Skills CLI (Vercel)
npx skills add shinpr/claude-code-workflows --skill requirement-convergence -a gemini-cli2Platform CLI
gemini skills install https://github.com/shinpr/claude-code-workflows --path dev-skills/skills/requirement-convergenceManually: put the files in place
.devin/skills/requirement-convergence/SKILL.mdThe legacy .windsurf/skills/ path still works
Claude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about requirement-convergence
How to install requirement-convergence in Claude Code?
Run npx skills add shinpr/claude-code-workflows --skill requirement-convergence -a claude-code or place the file directly at ~/.claude/skills/requirement-convergence/SKILL.md.
What does requirement-convergence do?
It verifies scope before design starts by isolating a single observable outcome, tracking non-goals, and evaluating repository evidence.
When is a convergence record considered complete?
Convergence completes when every applicable field reaches the ready readiness label, or when the user explicitly agrees to leave an unresolved field as weak-but-explicit.
Source text
SKILL.md · MIT
--- name: requirement-convergence description: Separates the outcome a change must produce from the requirements proposed to reach it, records what the user excluded, and bands cost from decision-relevant repository evidence. Use when a requirement enters a workflow, before design begins. --- # 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 | Field | Pass condition | |-------|----------------| | `outcome` | One 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. | | `cost` | A band with the repository and lifecycle 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, not from 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 a 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](references/criteria.md). ## Hearing Protocol Run the hearing after scope analysis has produced the facts needed to judge the convergence fields. The workflow using this skill owns the interaction method and routing; this skill defines the hearing content and pass conditions. Register these steps before starting and record each step's evidence as it completes: | Step | Action | Completion evidence | |------|--------|---------------------| | 1 | Render the Scope Confirmation below, asking only about the fields below `ready` | Each fact cites the analysis output it came from, and **User decisions** holds one question per field below `ready` | | 2 | Record each answer as that field's value | The value uses wording the user supplied, not wording the hearing offered | | 3 | Re-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 stands | Two recorded answers, or the user's agreement to stop | | 4 | Hand the record to the step that judges the fields | An updated record returned from that step | Step 2's evidence is what keeps the hearing reviewable: a value restating the hearing's own candidates fails it, so the user's judgment survives however the question was put. ## Scope Confirmation Render this shape at every requirements confirmation stop, whether or not the hearing ran, using only what can change the user's requirement decision or the workflow route. | Section | Contents | |---------|----------| | **User-selected scope** | The requirements and exclusions the user has already selected, in the user's wording | | **Decision evidence** | Each material observed fact with its source, followed by what that fact can change about scope, outcome, or cost | | **User decisions** | One question per convergence field below `ready` (Hearing Protocol Step 1), followed by the scope, outcome, or cost effects of its materially different answers | | **Workflow** | Rough cost band with the unknowns that remain, Structural Scale, and the selected document and workflow route | Keeping the first three sections separate is what makes the decision informable: merged, the reader cannot tell which line is the user's own selection, which is a repository observation, and which is still open. Only an explicit user answer moves an item from **User decisions** into **User-selected scope**. Because Step 2 records the user's own wording, pose each open decision as a question and let the user supply its answer. The orchestrator owns **Workflow** and presents it as a selected route. ## Storage Protocol | Carrier | Holds | Written by | |---------|-------|------------| | The convergence record in the judging step's output | Every field with its readiness label | The judging step | | PRD `Success Criteria` and `Future / Out of Scope` | `outcome`; user-authored `nonGoals` | The PRD production step | | Design Doc `Requirement Convergence` | The same when no PRD exists, and the fields left `weak-but-explicit` in every case | The Design Doc production step | A flow that produces neither document carries the record in its own context to the next step. ## Reference Protocol (For Downstream Consumers) 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 downstream 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. When work depends on it, return the missing decision and its effect to the owning workflow. ## 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](references/criteria.md) — judgment rules per field, cost inputs, challenge intensity, solution-in-disguise test
© shinpr · MIT · shinpr/claude-code-workflows