swarm

Dumky's rigorous engineering mode for non-trivial work.

Spawn four parallel review subagents over the active transcript, surface learnings, and route each to a concrete edit on an existing skill.

The twenty-two engineering principles dstack works from, one rule each, grouped as core, architecture, verification, delegation, and meta.

Strip leaked edit history from an existing artifact: v2 names, 'now supports' docs, '(was X)' UI copy, comments that describe the change instead of the code.

Write or edit a SKILL.md that actually changes agent behavior, and package it so both Claude Code and Codex load it.

Use for "interrogate", "adversarial review", "multi-model review", "challenge this", "stress test this code", "review this plan", "attack this plan", "find…

Sketch types, signatures, and module structure before code, then stay in the loop while implementation fills in.

Configure which model dstack uses per role.

Spawn comment-sicko, fix accepted findings, and offer encodings for claimed constraints.

Generate a project-local verification skill that drives your app the way a user does — any language, framework, or platform.

Design an auditable playbook when no narrower one fits: a large migration, an ambitious multi-part change, or work a human reviews after stepping away.

Spawn N parallel candidates at the same task, pick a base, graft the strongest parts of the losers into it.