gardeners
Autonomous project gardening by a coordinated team of agents.
Skill by maragudk
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/maragudk-gardeners
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: productivity
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md and code. Checked Oct 26, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/gardeners
What's inside
- Skill name
gardeners- Description
- Autonomous project gardening by a coordinated team of agents. Spawns a team of gardeners that each run the `garden` skill in parallel, coordinating via a shared task list to avoid duplicate work. Use when the user wants to tend multiple small issues in one pass. Invoke with /gardeners.
- License
- MIT
- Instructions
- 95 lines in SKILL.md · view on GitHub
Files (1)
- SKILL.md
SKILL.md (first lines)
# Gardeners A team version of the `garden` skill. Instead of one gardener pulling one weed, you spawn a small team that each pulls a different weed in parallel. They share a task list so two gardeners don't fight over the same issue. Use this when the user wants a broader sweep than a single `garden` run would do -- several small, independent issues fixed in one pass. For a single focused fix, use the `garden` skill directly instead. ## Flow 1. **Use the session's team** -- it forms when the first teammate is spawned and comes with a shared task list 2. **Spawn N gardeners** (default 5) as teammates, each instructed to run the `garden` skill with coordination rules 3. **Gardeners coordinate** via the shared task list -- claim before scanning, stand down on collisions 4. **Collect results** as each gardener reports in with a PR URL 5. **Review and merge** the PRs, then clean up ## Step 1: Use the session's team There's no team to create: each session has a single implicit team with a shared task list, formed automatically when the first teammate is spawned. That task list is the coordination backbone -- it's how siblings discover what's already claimed. ## Step 2: Spawn gardeners Default to **five gardeners** unless the user asks for a different count. More gardeners means more coverage but more collisions; fewer means less parallelism. Spawn each one with the `Agent` tool, passing a distinct `name` and `model: "opus"` -- gardening tasks are small and well-scoped, so they don't need the most expensive model. Sub-agents a gardener spawns (like its self-reviewers) inherit the model automatically. Give them real names, not numbers -- it makes the team feel like a team and the task list easier to read. Pick from the default roster of gardener names and extend it if you need more than five: - Herb - Basil - Rose - Fern - Pip - Sage - Hazel - Rosemary Give each the same prompt -- the coordination rules are what keep them from stepping on each other. ### Gardener prompt template Each gardener needs: - An identity (their name in the team) - A reminder that siblings exist - The coordination rules below - Instructions to run the `garden` skill to completion and report back the PR URL The coordination rules (include these verbatim in every gardener prompt): ``` 1. Before scanning, call TaskList to see what other gardeners have already claimed or completed. Avoid duplicating their work. 2. When you pick an issue, immediately call TaskCreate with a specific subject (e.g. "Fix typo X in file Y") and TaskUpdate to set yourself as owner and status=in_progress. This tells siblings what you're working on. 3. If another gardener has already claimed the issue you'd have picked, scan for a different one -- don't open a duplicate PR. 4. On collision (two gardeners on the same issue), earliest claim wins. The later claimer stands down and picks something else. 5. When done, TaskUpdate status=completed and include the PR URL. 6. You fix exactly one issue. If the scan surfaces others, leave them for the next gardening pass -- don't spin up more tasks or expand scope. ``` Run gardeners in the background (`run_in_background: true`) so they work in parallel. You'll get a notification as each reports in. ## Step 3: Let them work Gardeners will: - Post a claim to the task list when they pick an issue - Broadcast (via `SendMessage`) if they detect a collision, citing earliest-claim-wins - Open a PR and mark their task completed You generally don't need to intervene. If a gardener seems stuck, send them a message or check the task list for blocked tasks. ## Step 4: Collect results As each gardener reports in, note the PR URL. Expect between `N-2` and `N` PRs from a team of `N` -- some gardeners may find nothing new to pick after coordination, which is fine. ## Step 5: Merge and clean up Review the PRs together. Look for: - Duplicates that slipped through (close the losers) - PRs that should be combined Then: - Merge the good ones (squash, delete branch) - Close any duplicates with a comment pointing at the survivor - Prune stale local `garden/*` branches that are left over from the team run - Shut down the gardeners: `SendMessage` a `shutdown_request` to each one. The team itself needs no teardown -- its shared state is cleaned up automatically when the session ends ## Notes - **Earliest claim wins** is the simplest collision-resolution rule and the one that worked in practice. Don't overthink it. - **Shared task list beats chat** for coordination. Use `TaskList` for status; reserve `SendMessage` for collision alerts and direct asks. - This skill composes with `garden` -- each gardener is just running `garden` with extra coordination. If the single-gardener flow changes, this one benefits automatically. - **Talk like a gardener -- to the user.** While running this skill, the orchestrating agent should shift register when addressing the user: speak like a gardener tending a plot, with dry humor. Weeds, flower beds, pruning, seasons. Keep it subtle -- one turn of phrase per update, not a costume. Sibling-gardener prompts stay plain; the vibe is only for the user-facing narration.
From maragudk/fabrik, MIT.
More from maragudk/fabrik
- 0.652score
ableton-live
maragudk
Guide for making music in Ableton Live 12 on macOS.
SkillAFreshness: A★ 30Claude Code +9 - 0.616score
macos-design
maragudk
Design and build native-feeling macOS application UIs.
SkillBFreshness: B★ 30Claude Code +9 - 0.616score
security-review
maragudk
Perform a thorough security review of the project, starting from a randomly selected file.
SkillBFreshness: B★ 30Claude Code +9 - 0.616score
swiftui-pro
Paul Hudson
Comprehensively reviews SwiftUI code for best practices on modern APIs, maintainability, and performance.
SkillBFreshness: B★ 30Claude Code +9 - 0.616score
simple-english
maragudk
Write or rewrite technical text with the rules of ASD-STE100 Simplified Technical English so it is clear, unambiguous, and free of AI slop.
SkillBFreshness: B★ 30Claude Code +9 - 0.616score
observability
maragudk
Guide for instrumenting and operating observable software systems.
SkillBFreshness: B★ 30Claude Code +9 - 0.616score
xbrl
maragudk
Guide for working with XBRL, the standard for digital business and financial reporting -- reading and writing instance documents, Inline XBRL (iXBRL)…
SkillBFreshness: B★ 30Claude Code +9 - 0.611score
distill-book
maragudk
Distill a long book into a concise, structured set of learnings by processing it chapter by chapter with parallel subagents, then synthesizing the result --…
SkillBFreshness: B★ 30Claude Code +9 - 0.611score
decisions
maragudk
Guide for recording high-level architectural, design, and product decisions in docs/decisions.md.
SkillBFreshness: B★ 30Claude Code +9 - 0.611score
address-code-review
maragudk
Address code review feedback by walking through comments one at a time with the user.
SkillBFreshness: B★ 30Claude Code +9 - 0.611score
garden
maragudk
Autonomous project gardening.
SkillBFreshness: B★ 30Claude Code +9 - 0.611score
improve-skill
maragudk
Review the current conversation for fabrik skills that could be improved (corrections, friction the user had to manually flag, missed triggers, anything else…
SkillBFreshness: B★ 30Claude Code +9
Signals
- Stars
- 30
- Forks
- 2
- Open issues
- 0
- Last commit (this folder)
- Jul 28, 2026 (3 months ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/maragudk/fabrik
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from maragudk
- 0.652score
ableton-live
maragudk
Guide for making music in Ableton Live 12 on macOS.
SkillAFreshness: A★ 30Claude Code +9 - 0.616score
security-review
maragudk
Perform a thorough security review of the project, starting from a randomly selected file.
SkillBFreshness: B★ 30Claude Code +9 - 0.616score
xbrl
maragudk
Guide for working with XBRL, the standard for digital business and financial reporting -- reading and writing instance documents, Inline XBRL (iXBRL)…
SkillBFreshness: B★ 30Claude Code +9 - 0.616score
macos-design
maragudk
Design and build native-feeling macOS application UIs.
SkillBFreshness: B★ 30Claude Code +9 - 0.616score
observability
maragudk
Guide for instrumenting and operating observable software systems.
SkillBFreshness: B★ 30Claude Code +9 - 0.616score
simple-english
maragudk
Write or rewrite technical text with the rules of ASD-STE100 Simplified Technical English so it is clear, unambiguous, and free of AI slop.
SkillBFreshness: B★ 30Claude Code +9
Cite this page
gardeners skill. Agent Mart. https://agentmart.dev/skills/maragudk-gardeners (last updated 2026-10-24)