agentsclimarketplace

Linear seed

Skill Manzanita-Research/magpie/plugins/linear/skills/linear-seed

πŸ¦β€β¬› Opinionated agentic project management built on Linear. For solo devs and small teams with too many repos.

Install
npx -y skills add Manzanita-Research/magpie --skill linear-seed

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 4 stars4 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Capture project ideas as lightweight Linear issues before they become real repos. Use when the user floats an idea, riffs on a concept, says "what if we...", "we should build...", "imagine if...", "idea:", or any variation of sharing a project idea they're not ready to commit to. Also triggers on "catch this idea", "pin this thought", "seed this", or "add to the wall". This is the sticky-note-on-the-studio-wall skill.

SKILL.md

5.2 KB, as published. Nobody here has run it

Seed

You collect ideas like a magpie. This skill catches them before they fly away.

When someone riffs on an idea β€” "what if we made a Spotify automation tool for music communities" β€” capture it as a Seed in Linear. Not an issue to work on. Not a project to plan. Just a sticky note on the wall.

What a Seed is

A Seed is a Linear issue with the Seed label. That's it. No Product label (it doesn't have a repo yet), no project, no priority. Just a title, a description, and enough context to remember why you were excited about it.

Seeds start in Seeds (raw ideas, waiting to germinate) until one of three things happens:

  1. It grows β€” you create a repo, add a Product label, remove the Seed label, and start building
  2. It merges β€” the idea gets folded into an existing project as a feature
  3. It composts β€” you move it to Composted. Not a failure β€” fertilizer. Most seeds don't sprout.

Catching ideas

When the user floats an idea, draft a seed from context. Don't interrogate them β€” the whole point is low friction.

Title: The idea in one line, the way you'd text it to a friend.

  • "spotify automation tool for irl music communities"
  • "visual EQ that maps frequency bands to colors"
  • "cli tool that turns voice memos into commit messages"

Description: A short paragraph capturing the core of the idea. What is it, who is it for, why does it matter. If the user riffed for a while, distill it β€” don't transcribe.

Don't set:

  • Priority (it's not urgent, it's an idea)
  • Project (it doesn't belong to one yet)
  • Product label (it doesn't have a repo yet)
  • Assignee (nobody's working on it)

Confirm before filing

Always show the user what you're about to file:

Seed: spotify automation tool for irl music communities

Release Radar but for your people. Maintain a config of artists in your community, auto-fetch new releases, build playlists. No algorithmic curation β€” just what your scene is making.

File this?

Keep it fast. One confirmation, not a form.

Execution

After the user confirms, delegate the Linear work to a background subagent. This keeps the main context clean β€” seeding an idea shouldn't eat 10 tool calls in your conversation.

  1. Use the Agent tool with subagent_type="general-purpose" and run_in_background=true
  2. Include in the prompt:
    • The full text of plugins/linear/LINCTL_REFERENCE.md (read it first if not already in context)
    • The seed title and description
    • Team key if already known from session context
    • If the user riffed longer, include the comment body to add after creation
  3. The subagent should:
    • Run linctl whoami to verify auth
    • Run linctl team list --json to get the team key (if not provided)
    • Create the issue with linctl issue create --title "..." --team <KEY> --labels "Seed" --state "Seeds" --description "..."
    • If there's a longer riff, add it as a comment with linctl comment create <ID> --body "..."
  4. Report back to the user with the issue identifier and a confirmation

See plugins/linear/LINCTL_REFERENCE.md for exact command syntax and gotchas.

When the user riffs longer

Sometimes an idea comes with context β€” "and it could work like X, and we'd need Y, and the hard part is Z." When the conversation has more substance than fits in a description:

  1. File the seed with a concise description (the elevator pitch)
  2. Have the subagent add a comment with the longer riff β€” the context, the technical ideas, the open questions

This way the seed stays scannable in list view but the full context is preserved.

When NOT to file a seed

  • The user is describing work on an existing project β€” use linear-workflow instead
  • The user is thinking out loud but hasn't landed on anything ("I wonder if..." followed by "nah, never mind")
  • The user explicitly says they don't want to track it

If you're unsure: "Want me to seed this idea in Linear?"

Reviewing seeds

When the user asks what's on the wall, what ideas they've had, or wants to review seeds:

linctl issue list --team <TEAM_KEY> --state "Seeds" --json

Present them as a simple list β€” title and creation date. Offer to:

  • Promote one to a real project (create repo, add Product label, remove Seed)
  • Enrich one with more context (add a comment)
  • Compost ones that aren't going anywhere (move to Composted state)

Promoting a seed

When a seed is ready to become a real project:

  1. Ask what the repo name should be (suggest one based on the title)
  2. Remove the Seed label
  3. Add the appropriate Product label (create one if needed)
  4. Optionally assign it to a project
  5. The issue keeps its history β€” the original idea is still there

Don't auto-create a GitHub repo. That's a separate decision. Just update the Linear issue.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.