linear-requirement
/linear-requirement
Pisze i zgłasza po zatwierdzeniu zgłoszenie wymagań Linear (## Requirement) dla Sokosumi, tylko na wyraźną prośbę o zgłoszenie ticketu.
- Autor
- masumi-network ↗
- Repozytorium
- sokosumi ↗
- Licencja
- MIT
- Ostatnia zmiana repozytorium
- 26.10.2026
Ocena i bezpieczeństwo
- Pliki w paczce
- 5
- Skrypty
- 0
- Komentarze HTML
- 0
- Adresy URL w SKILL.md
- 1
Brak sygnałów. Repozytorium zmienione 26.10.2026, bez odwołań do wycofanych modeli i API.
MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.
Co robi i kiedy
Pisze i zgłasza po zatwierdzeniu zgłoszenie wymagań Linear (## Requirement) dla Sokosumi, tylko na wyraźną prośbę o zgłoszenie ticketu.
Wyzwalacze z opisu
Oryginalny opis (EN)
Draft and file a Sokosumi Linear requirement issue (## Requirement) after user approval. Use only when the user explicitly asks to file/draft a Linear requirement, ticket, or SOK issue (e.g. "file this as a Linear issue", "draft a requirement", "create a SOK ticket", /linear-requirement). Never use under implement, bug-fix, feature, refactor, PR review, or casual chat about a bug/feature when the user did not ask to file a ticket.
Jak działa
- Runtime compatibility
- Defaults
- Workflow
- Writing style
- Label classification
- Gold-standard requirement reference
- Supporting files
Podgląd SKILL.md
---
name: linear-requirement
description: >-
Draft and file a Sokosumi Linear requirement issue (## Requirement) after user
approval. Use only when the user explicitly asks to file/draft a Linear
requirement, ticket, or SOK issue (e.g. "file this as a Linear issue",
"draft a requirement", "create a SOK ticket", /linear-requirement). Never use
under implement, bug-fix, feature, refactor, PR review, or
casual chat about a bug/feature when the user did not ask to file a ticket.
disable-model-invocation: true
---
# linear-requirement
You are the **requirement agent**. Turn a rough feature, bug, or improvement into a concise **Linear issue** with a `## Requirement` section — enough to start work later, not a final spec.
**Explicit ask only.** Do not start this workflow because implement work or a Spec mentioned a "requirement". The user must ask to draft or file a Linear issue (or invoke this skill).
**Approval gate:** Show the draft in chat and **wait for explicit user approval** before any Linear write.
After the issue exists (or is updated), stop. Do not start `/to-spec` or `/implement` in this session. Only touch the requirement issue fields listed in `LINEAR-CLI.md` (or `LINEAR-MCP.md` when the CLI is missing) — no other Linear side effects.
```mermaid
flowchart LR
user["User describes\nfeature / bug / improvement"] --> intake["Intake +\ndupe search"]
intake --> draft["Draft requirement\n(chat only)"]
draft --> approve{"User\napproves?"}
approve -->|no| draft
approve -->|yes| issue["Create or update\nLinear issue"]
```
## Runtime compatibility
Canonical files: `skills/linear-requirement/`. Invoke by name or explicit ask (`disable-model-invocation`). Install into agent skill dirs with `npx skills add . --skill linear-requirement` (or `npx skills add masumi-network/sokosumi --skill linear-requirement` from another clone).
## Defaults
| Field | Value |
|-------|-------|
| Linear team | `SOK` |
| Linear project | `sokosumi-6357694ddd23` (Sōkosumi) |
| Linear state | `Triage` |
| Linear priority | `3` (Medium) unless user overrides |
| Linear assignee | unassigned (omit `--assignee`) unless user overrides |
| Linear label | Infer exactly one: `Feature`, `Bug`, or `Improvement` |
| Title | Plain product title (not Conventional Commit) |
Do not ask for the Linear project by default.
## Workflow
1. **Intake**
- Required: feature summary, bug report, or improvement idea (plain language).
- Optional: priority, assignee, locked decisions, label override, project override, existing issue id to refine.
- Track **publish target**: `create` (default) or `update:SOK-XXX`.
- If intake is vague, ask **one** clarifying question and wait.
- If intake names an existing Linear issue to refine, load it (`linear issue view SOK-XXX --json --no-pager --no-download`, or `get_issue` if `linear` is missing), set publish target to `update:SOK-XXX`, and treat the body as raw input — not approved yet.
2. **Light discovery + duplicate check**
- Search only what you need to ground the requirement: related routes, services, schemas, or UI areas.
- Read scoped `AGENTS.md` when the area is clear.
- If publish target is still `create`, run a **duplicate search** before drafting:
1. Build 2–5 keywords from the problem (product area + symptom/goal).
2. CLI: `linear issue query --team SOK --project sokosumi-6357694ddd23 --search "<keywords>" --limit 5 --json --no-pager`. If `linear` is missing: `list_issues` filtered to team `SOK` / project Sōkosumi.
3. Show up to **5** closest hits (id, title, state, URL).
4. Treat as a **close match** when same product area **and** same user problem (not merely the same feature area).
5. If any close match exists, ask: **update that issue**, **file a new one**, or **stop**.
6. If the user picks an issue to update, set publish target to `update:SOK-XXX` immediately and load it (`linear issue view` or `get_issue`) before drafting.
- Resolve obvious product/scope questions from code — do not invent file-level plans.
- Keep notes short. No research report.
3. **Draft requirement**
- Fill `REQUIREMENT-TEMPLATE.md`.
- Infer label and a **plain product title** (what a human scanning Linear should understand). Do **not** use Conventional Commit prefixes (`feat:`, `fix:`) for the Linear title.
- Show near the top (chat draft only — do **not** post this line to Linear). For **create**, use defaults. For **update**, show the existing issue’s current state/assignee/priority unless the user overrode them:
```markdown
**Requirement draft:** create · project Sōkosumi · state Triage · priority Medium · assignee none · label Feature
```
```markdown
**Requirement draft:** update SOK-XXX · project Sōkosumi · state In Progress (keep) · priority Medium · assignee (keep) · label Bug
```
- Do **not** include: file lists, contract tables, verification commands, mermaid data-flow diagrams, or coder breakdown. Those belong in a later design/spec step if someone builds the issue.
4. **Approval gate (required)**
- Present the full draft in chat under a clear heading, e.g. `## Draft requirement — review before Linear`.
- Include proposed title, label, **publish target** (`create` vs `update SOK-XXX`), and the requirement body.
- End with:
```text
Reply **approve** to post this Linear requirement.
Reply with edits to revise the draft.
```
- **Stop.** Do not run Linear writes (`linear` or MCP) until the user approves.
- On edits: revise and show the draft again. Repeat until approved.
5. **Publish (only after approval)**
- If `command -v linear` succeeds, read `LINEAR-CLI.md` and run its health check. If `linear` is missing, read `LINEAR-MCP.md` and run its MCP health check.
- **Create** (publish target `create`): CLI `linear issue create` or MCP `save_issue` with **all** required create fields — always include project `sokosumi-6357694ddd23` when the user did not override it (do not skip it; do not use `"Sokosumi"`).
- **Update** (publish target `update:SOK-XXX`): follow **Update existing issue** in the file you opened — merge `## Requirement`, optional title/labels (preserve non-type labels); **do not** reset `state` to create defaults.
- After write, run post-write verify in that same file.
- Return issue id/URL, label, project, assignee, priority, and state.
- If both `linear` and Linear MCP are unavailable, return the approved draft and say so. Do not use browser automation, curl, or Linear REST.
## Writing style
- Straight to the point. Problem and goal in one or two sentences each.
- Bullets over paragraphs.
- Cite real file paths or areas only when they clarify scope — not as an implementation plan.
- Keep **Out of scope** explicit.
- Do not over-specify architecture; leave design details for a later step.
## Label classification
| Label | Use when |
|-------|----------|
| `Feature` | New user-facing capability, new route/page/API, greenfield behavior |
| `Bug` | Broken or incorrect existing behavior, regression, crash, wrong data |
| `Improvement` | Enhancement to something that already works: UX, perf, refactor, DX, a11y |
Decision order:
1. User says bug / broken / fix / regression → `Bug`.
2. User says improve / polish / optimize / refactor and no new capability → `Improvement`.
3. New capability users did not have → `Feature`.
4. Only changes an existing flow → `Improvement`.
5. If unclear, default `Feature` for greenfield, `Improvement` for iteration on shipped code.
## Gold-standard requirement reference
[SOK-537](https://linear.app/masumi/issue/SOK-537/create-history-view) — problem, goal, decisions, rough architecture, references, out of scope. No files, no verification.
## Supporting files
- `REQUIREMENT-TEMPLATE.md` — requirement body shape.
- `LINEAR-CLI.md` — create/update after approval when `linear` is on PATH.
- `LINEAR-MCP.md` — create/update after approval when `linear` is missing.
- `.agents/skills/linear-cli/` — CLI flags and recipes.
Pliki w paczce (5)
- AGENTS.md 1 KB
- LINEAR-CLI.md 7 KB
- LINEAR-MCP.md 7 KB
- REQUIREMENT-TEMPLATE.md 1 KB
- SKILL.md 8 KB
Wersje
W bazie nie ma innych skilli o nazwie linear-requirement.
Migawka bazy z 13.11.2026 · wersja paczki 347197e211e5 (skrót SHA-256 całej paczki)
Pobrania w ostatnich 30 dniach: 0 · ranking
Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.
Podobne skille
- 100linear-ticket
Zakłada w Linear zgłoszenie z błędu, funkcji, drobnego zadania lub usprawnienia, gdy użytkownik prosi o ticket lub issue. - 90issue
Przygotowuje zgłoszenie w Linear, które programista lub agent podejmie bez dodatkowych pytań, po zatwierdzeniu podglądu przez użytkownika. - 100linear-issue
Pobiera zgłoszenie Linear wraz z powiązanym kontekstem i analizuje je przed rozpoczęciem pracy nad zadaniem. - 100pr-link-issue
Dopisuje do opisu bieżącego PR link do zgłoszenia GitHub i powiązane zgłoszenie Linear, ustalane automatycznie z komentarza linear-linkback. - 100linear-workspace
Zarządza zgłoszeniami Linear i kieruje pracę do wyspecjalizowanych agentów oraz komend: tworzenie, aktualizacja, wyszukiwanie, edycja zbiorcza, porządkowanie i synchronizacja. - 100linear-ticket-lifecycle
Znajduje lub tworzy zgłoszenie Linear przy każdej zmianie kodu, przenosi je do In Progress na starcie i do In Review po zakończeniu, żeby praca była śledzona.