skills.rest
agentculture
agentcultureOFFICIAL
Explore all skills by this author →
WebsiteGitHub

assign-to-workforce

Assign dependency-wave tasks to parallel agents in isolated git worktrees.

112|6|Updated Mar 19, 2026
Search Tags:#tdd#implementation planning#git worktrees#parallel tasks#agentic orchestration#devague
One-click install
npx skills add https://github.com/agentculture/culture --skill assign-to-workforce
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: assign-to-workforce
Source: https://github.com/agentculture/culture/tree/main/.claude/skills/assign-to-workforce
Command: npx skills add https://github.com/agentculture/culture --skill assign-to-workforce

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires bash, git, python3, uv, devague, and includes scripts (resource) components.

What problem does it solve?

It prevents slow, sequential implementation work by turning a converged dependency graph into parallel, test-driven task execution across isolated worktrees, while keeping human approval confined to the intended gates.

Core Features & Use Cases

  • Dependency-wave fan-out: Converts devague plan waves scheduling metadata into a human-reviewable implementation split plan.
  • Isolated worktrees per task: Assigns one agent per task per wave in separate git worktrees to avoid clobbering shared files.
  • TDD-gated reconciliation: Ensures every task merge is validated by running tests before and after merging into the main branch.
  • Three human gates only: Requires approval for the exported spec, the implementation split plan (go/no-go), and the final PR—no per-task human pauses.

Quick Start

Ask your AI operator to run assign-to-workforce split-plan for the current plan, review the proposed table, then approve to fan out the waves for TDD-gated merges.

Frequently Asked Questions about assign-to-workforce

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run parallel implementation tasks from a devague plan without file conflicts?▼

Parallel implementation tasks from a devague plan run in isolated git worktrees to prevent file clobbering. The system assigns one agent per task per dependency wave, keeping concurrent work separated until TDD-gated merges.

What is dependency-wave fan-out in agentic orchestration?▼

Dependency-wave fan-out converts devague plan scheduling metadata into a human-reviewable implementation split plan. It groups tasks by dependency waves so parallel agents can execute them concurrently across isolated git worktrees.

Do I need to manually approve every parallel task during agent orchestration?▼

No, agent orchestration requires human approval at only three gates: the exported spec, the implementation split plan, and the final PR. Individual parallel tasks run without per-task human pauses until TDD-gated reconciliation.

How are parallel task merges validated when fanning out plan waves?▼

Parallel task merges are validated through TDD-gated reconciliation, which runs tests before and after merging each task into the main branch. This ensures isolated worktree changes do not break the shared codebase.

Can I use devague CLI orchestration inside isolated git worktrees for subtasks?▼

No, strict rules prevent devague CLI orchestration inside task worktrees. This avoids recursive agent spawning and ensures parallel tasks remain isolated until the human go/no-go checkpoint and final PR merge.

Related Skills

Explore More Software Engineering Skills→
agentcultureagentculture
112

spec-to-plan

Convert converged spec frames into buildable task plans with acceptance criteria.

Official
Intermediate
YanCheng-goYanCheng-go
3

commit

Automate branch creation, commits, PR review, and gated merge with human approval.

Community
Advanced
son-nguyen-301son-nguyen-301

split-plan

Decompose parent plans into parallel subtasks and orchestrate execution in dependency waves.

Community
Advanced
26mitch2626mitch26

dispatching-parallel-agents

Dispatch independent tasks to specialized agents for parallel execution without shared state.

Community
Intermediate