swarm
Fan out parallel subagents and return one report.
Source facts
- Repository
- Belweave/hstack
- Last source activity
- October 28, 2026 at 07:54
- Detected SKILL.md language
- English
- Stars
- 0
- Forks
- 0
Install options
The review-first prompt is selected by default. You can switch to a direct command or download a local copy.
Review the source files
Read SKILL.md and any companion files shown by SkillsMP before deciding whether to install.
Showing SKILL.md
SKILL.md
Source instructions · Read-only preview- name
- swarm
- description
- Fan out parallel subagents and return one report.
- version
- 0.1.0
- author
- Preetham Kyanam
- license
- MIT
- platforms
- ["linux","macos","windows"]
- metadata
- {"hermes":{"tags":["delegation"],"related_skills":["prove"]}}
# Swarm
## When to Use
The task splits into slices that do not write the same files. This is not `arena`. Swarm partitions the work. Arena duplicates the same brief.
## What to do
Send one `delegate_task` batch. Stay inside `delegation.max_concurrent_children`. That cap is 10 unless the user raised it. A larger batch is an error, not something to work around.
Each child gets the goal, its slice, and the check that slice must run. Children do not watch pull requests and do not spawn their own panels.
Races are allowed when the slices are competing proofs of the same question.
Return one report. Keep the findings, not each child's transcript. Load `prove` for the check each slice owes.
View on GitHub