Worker Delegation Playbook
Teaches an agent to split colony work into parallel worker jobs tracked in a table
- Repository
- aden-hive/hive
- Stars
- 11,132
- Licence
- Apache-2.0
- Files
- 1 file
What it does
This is an operating guide for an AI agent working in the Hive framework's colony mode. It explains when it is worth splitting a large job (such as many rows, accounts, or files) across several parallel worker agents, and when doing it yourself is better. It tells the agent to track each unit of work as a row in a table, write a reusable worker procedure, test that procedure on one row first, and then run a Python playbook that keeps dispatching workers until every row is marked done, so an interrupted run can resume.
Great for
- Developers building or running multi-agent workflows on the Hive framework
- People who want an agent to process many similar items (for example, profiles or records) in parallel and resume if it stops partway
Skip it if
- Simple or exploratory tasks with only one or two items, since the guide itself says spawning workers has overhead that outweighs the benefit
- Setups without the Hive colony tools (run_playbook, tracker_sql, write_skill), where the instructions cannot be followed as written
Try asking
- “Fan out enriching 300 rows in my prospect tracker across parallel workers, testing one row first”
- “Set up a tracker table and playbook so a batch scrape of 80 accounts can resume if it fails halfway”
- “Decide whether splitting these 5 research files across workers is worth it or if I should do them myself”
Install in 3 steps
- 1
Download the skill
The ZIP contains the 1 original files, the Apache-2.0 licence and an attribution note.
worker-delegation.zip - 2
Add it to your agent
Claude Code (all projects)
mkdir -p ~/.claude/skills && unzip worker-delegation.zip -d ~/.claude/skills/Claude Code / Agent SDK (this project only)
mkdir -p .claude/skills && unzip worker-delegation.zip -d .claude/skills/Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.
- 3
Ask for the task
No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.
Get better results
- Always pilot one row yourself before starting parallel workers, especially on sites with dynamic pages or login, as the guide recommends
- Include a status or timestamp column in the tracker so the playbook knows which rows are still undone
- Ask the user before fanning out if you are unsure whether a shared browser session or login can safely be used by several workers at once
What's inside
- SKILL.md
Redistributed unmodified from aden-hive/hive under the Apache-2.0 licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.