Research Backlog

research-backlog

A Claude Code skill that enriches Linear backlog tickets with deep research before a developer touches them.

What it does

For each ticket in your Linear team's backlog, dispatches a parallel research subagent that investigates the technical domain, searches for official docs, prior art, and best practices, then posts a structured comment directly on the ticket. The result is a backlog where every item arrives at the developer with context, resources, and known gotchas already attached — reducing ramp-up time and design mistakes.

Installation

Copy the skill folder into your Claude Code skills directory:

cp -r research-backlog ~/.claude/skills/

Then invoke it in Claude Code with `/research-backlog`.

Usage

Run `/research-backlog` from any repo. The skill will:

1. Confirm your LINEAR_API_KEY
2. List your teams (or auto-select via $LINEAR_TEAM_NAME)
3. Fetch all backlog/unstarted issues
4. Dispatch one research subagent per ticket in parallel
5. Post a structured research comment on each ticket
6. Print a summary table of what was posted

Requirements

• LINEAR_API_KEY — required. Get one from linear.app/settings/api.
• LINEAR_TEAM_NAME — optional. If set, skips the team selection prompt. Value should match part of your team's name (case-insensitive).
• jq — used to parse GraphQL responses (brew install jq)

License

MIT