parallel-work

A workflow for splitting GitHub issues among several Claude Code agents, with each agent working in its own isolated Git worktree, a separate checkout of the repository.

In plain words
What is it for?
Use it to implement multiple non-conflicting GitHub issues at the same time and have agents commit their work and open pull requests.
Why use it?
It removes the manual work of selecting separate tasks, creating branches and worktrees, and gathering the agents' completed changes.

Skill for Claude CodeCodex

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add skills/destrayon/connapse/parallel-work
Any agent
npx skills add Destrayon/Connapse --skill parallel-work
Clone the repo
git clone --depth 1 https://github.com/Destrayon/Connapse

Made for: Claude Code, Codex.

Per session 76 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 2,567 The whole file, excluding the scripts and references it only reads on demand.
Security scan C 1 finding. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00076 $0.02567
Opus 5 $0.00038 $0.01283
Sonnet 5 $0.00015 $0.00513
Haiku 4.5 $0.00008 $0.00257

Measured 2d ago against content hash 7c5dd448eab9, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade C, and why

parallel-work scanned grade C with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Recursive force deletehighDestructive command

rm -rf with a variable or a broad path is one typo away from removing the wrong tree.

- **Worktree cleanup**: After all PRs are merged, remind the user they can clean up leftover worktrees with `rm -rf .claude/worktrees/agent-*`. Worktrees with no changes are auto-cleaned, but ones with commits persist.
.claude/skills/parallel-work/SKILL.md · 262 lines

How it starts

The opening of the file, as written. The whole thing — 262 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Parallel Work

Run N autonomous agents in parallel, each working on a separate GitHub issue in its own isolated worktree. Each agent creates a branch, implements the task, commits, and opens a PR.

Default: 3 agents. Override with argument: /parallel-work 2 or /parallel-work 5.

Parse the agent count from $ARGUMENTS if provided (extract the number). Default to 3 if not specified or not a number between 1-5.

Step 0: Pre-Flight Check

Before anything else, ensure a clean starting point:

  1. Check current branch: If not on main, warn the user. Worktrees branch from HEAD — spawning from a feature branch means agents won't have the latest merged code.
  2. Check for uncommitted changes: If the working tree is dirty, offer to stash (git stash -u -m "pre-parallel-work") before switching branches. Always restore the stash after switching.
  3. Switch to main and pull: git checkout main && git pull to ensure agents start from the latest code.
  4. Restore stash if needed: git stash pop to bring back any uncommitted work.

Only proceed to Step 1 after confirming the user is on an up-to-date main with their changes preserved.

Step 1: Gather Context

Run all of these in parallel:

Open issues

gh issue list --repo Destrayon/Connapse --state open --json number,title,labels,milestone,assignees,createdAt,updatedAt --limit 100

Recent merged PRs (for momentum scoring)

gh pr list --repo Destrayon/Connapse --state merged --json number,title,mergedAt,labels --limit 5

Current git state

git status --short
git branch --show-current

Project board state (filter to Todo only)

gh project item-list 3 --owner Destrayon --format json

Filter the results to items with status == "Todo" only. Exclude items with status "Done" or "In Progress" — those are already being worked on or completed.

Architecture context

Read .claude/state/decisions.md and .claude/state/issues.md for blockers or dependencies.

Read the full file on GitHub · 262 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 262 lines · 76 tokens per session scan C 7c5dd448eab9

Subscribe to this mod's changes

parallel-work is a skill published in the GitHub repository Destrayon/Connapse (16 stars, last pushed 3d ago), licensed MIT. It adds 76 tokens to every session and 2,567 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it C with 1 finding (recursive force delete). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Related

Other skills, from other repositories

cocosearch-onboarding

Use when onboarding to a new or unfamiliar codebase. Guides you through understanding architecture, key modules, and code patterns step-by-step using CocoSearch.

VioletCranberry/coco-search · 37 tokens

subcog-integrator

Analyze and enhance AI prompts, skills, commands, hooks, and system configurations to leverage Subcog's persistent memory system effectively.

zircote/subcog · 0 tokens

mcaf-solution-governance

Set up or refine solution-level governance for MCAF repositories: root and project-local AGENTS.md, rule precedence, solution topology, skill routing, and maintainability-limit policy placement. Use when bootstrapping a repo, restructuring a multi-project solution, or tightening agent rules.

managedcode/MCPGateway · 64 tokens

dotnet-microsoft-agent-framework

Build .NET AI agents and multi-agent workflows with Microsoft Agent Framework using the right agent type, threads, tools, workflows, hosting protocols, and enterprise guardrails.

managedcode/PrompterOne · 40 tokens

dotnet-mcaf

Adopt MCAF governance in a .NET repository with the right AGENTS.md layout, repo-native docs, skill installation, verification rules, and non-trivial task workflow. Use when bootstrapping or updating MCAF alongside the dotnet-skills catalog.

managedcode/PrompterOne · 57 tokens

producer

You are the producer agent. Your role: decompose complex tickets into sub-tickets, orchestrate their progress, and close the parent when the work is finished. You are the only agent that creates tickets.

Ekioo/KittyClaw · 0 tokens