🗺️ Agentic Methodology · Journey Playbooks

Playbook: Run parallel agent workstreams safely

Worktree isolation plus dispatch discipline letting multiple agents ship concurrently without collisions.

journey~85 minAgent Power UsersAI-Native EngineersTeams Shipping With Agents

Journey Steps

1Step 1 — Isolate parallel work with git worktrees: start with "Create worktrees per workstream: git worktree add ../feat-x -b feat-x"
2Step 2 — Dispatch subagents with fresh-context discipline: start with "Extract each task's brief to its own file; the brief is the ONLY requirements source"
3Step 3 — Receive agent code review without ego or obedience: start with "Reproduce or locate every claim before acting; reviewers are fallible too"
4How it fits together: One agent per worktree, one brief per task. Cross-tree edits are how weekends die.

Commands

Install all referenced skills

$ npx skills add aniruddhaadak80/skills --skill branch-isolation-git-worktree-isolation && npx skills add aniruddhaadak80/skills --skill execution-discipline-subagent-dispatch-discipline && npx skills add aniruddhaadak80/skills --skill branch-isolation-receiving-code-review-gracefully

Tags

#playbook#journey#agentic-methodology

Install this skill

$ npx skills add aniruddhaadak80/skills --skill playbook-agentic-methodology-parallel-workstreams -a opencode

SKILL.md

# Playbook: Run parallel agent workstreams safely

> Worktree isolation plus dispatch discipline letting multiple agents ship concurrently without collisions.

## When to Use This Skill

Worktree isolation plus dispatch discipline letting multiple agents ship concurrently without collisions.

## Steps

1. Step 1 — Isolate parallel work with git worktrees: start with "Create worktrees per workstream: git worktree add ../feat-x -b feat-x"
2. Step 2 — Dispatch subagents with fresh-context discipline: start with "Extract each task's brief to its own file; the brief is the ONLY requirements source"
3. Step 3 — Receive agent code review without ego or obedience: start with "Reproduce or locate every claim before acting; reviewers are fallible too"
4. How it fits together: One agent per worktree, one brief per task. Cross-tree edits are how weekends die.

View source on GitHub ↗

Take retrieval-augmented answers from empty repo to evaluated production feature.

🗺️ AI Engineering·~135m

Open →

🗺️Playbook: Harden an agent before giving it toolsjourney

Ship an autonomous agent whose failure modes are cheap, visible, and reversible.

🗺️ AI Engineering·~125m

Open →

🗺️Playbook: Cut LLM costs without quality collapsejourney

Reduce monthly inference spend measurably while keeping answer quality within tolerance.

🗺️ AI Engineering·~95m

Open →