| name | linear-issue-writer |
| description | Structure and draft Linear issue bodies, updates, and comments with strict anti-slop rules. Use when the user wants to write a Linear issue/update/comment or decide the shape of one; do not use for general Linear lookup, status checks, project browsing, or routine reads. |
Linear Issue Writer
Use this skill to avoid issue sprawl and bloated ticket bodies when writing Linear content. It does not replace linear:linear for ordinary Linear reads, lookup, status checks, or project management.
Default to one issue, draft first, and write to Linear only after explicit user approval.
Workflow
- Gather context.
- Read the user request and any linked issue, PR, doc, memory note, or code reference.
- If Linear tools are available, search/list nearby existing issues before proposing a new one.
- If the target team/project/status is unclear, ask only for the missing field needed to write the issue.
- Run the collapse pass.
- Decide:
create new, update existing, comment on existing, or do nothing.
- Prefer
update existing when the work shares the same deliverable, owner, validation path, or PR.
- Create separate issues only for separate owners, separate repos/PRs, separate validation paths, or independently shippable deliverables.
- Draft the issue or update.
- Show the collapse decision and the draft body.
- Keep one issue by default; use checkboxes inside
Done When for subwork.
- Ask for approval.
- Do not call Linear create/update/comment tools until the user approves the draft.
- After approval, create/update/comment exactly what was approved.
- Report the result.
- Return the issue key/link, what changed, and any fields left unset.
Issue Body
Use this structure unless the user asks for a different template:
## Goal
One sentence: what outcome this issue should produce.
## Context
- 2-4 bullets max.
- Link the existing issue/PR/doc/memory note instead of copying long background.
## Scope
- What will change.
## Non-goals
- What this issue will not do.
## Done When
- [ ] Observable completion criterion.
- [ ] Another criterion only if needed.
## Validation
- Smallest check, test, benchmark, log, or review evidence needed.
Anti-Slop Rules
- Title names an action or result, not a theme.
- No speculative follow-up work.
- No duplicated context from linked artifacts.
- No essay paragraphs; bullets should be short.
- No “investigate everything” issue unless the output artifact is named.
- No issue split just because there are multiple steps.
- No Linear write without saying whether it is create, update, or comment.
Collapse Output
Before drafting, show this tiny decision block:
Decision: create new | update existing | comment on existing | do nothing
Target: <new issue or existing issue key>
Why: <one sentence>
Split avoided: <what would have become extra issues, if any>
Linear Writes
- For
create new, create one issue with the approved title/body/team/project/status/labels.
- For
update existing, update only the approved fields; prefer appending a concise comment when changing the body would obscure history.
- For
comment on existing, write a status or scope comment using the approved draft.
- If approval changes the scope, revise the draft and ask again.
Fallback
If Linear tools are unavailable, produce the approved draft only and say that Linear write tools are not connected.