Automate Multi-Step PR Series with sesori-plan-worker

Where this fits

Developers often maintain a plan spanning multiple pull requests, each building on the previous one. When a PR merges, the next step should start automatically—manually re-entering the context is error-prone. This skill fits that workflow: it turns a static plan into an executable series that advances itself across merges.

Why agents benefit

• Agents can parse an existing PLAN.md or TRACKER.md and pick up the next logical PR without re‑asking for context.
• A monitored merge triggers the agent to automatically branch, implement, and open the successor PR, reducing round trips.
• The skill explicitly grants the agent authority to follow user instructions even when they deviate from the plan, so it doesn’t feel boxed in.
• It handles the bookkeeping: updating tracker files and step docs as work progresses, keeping durable truth in sync.
• It can work alongside a plan-maker skill, but doesn’t require one—it promotes anyone with a plan to executor.

Practical scenarios

Scenario 1: Implementing a stored plan from scratch
You have a PLAN.md that breaks a feature into three PRs. Ask the agent to “implement the plan” and it creates the first branch, writes code, tests, and opens PR #1. As soon as that merges, the agent auto‑starts PR #2, and so on.

Scenario 2: Resuming after a break or failure
A build failed mid‑series, and you’ve lost the thread. The skill reads TRACKER.md, sees which steps are incomplete, and proposes continuing from the last successful checkpoint—no lengthy re‑explanation.

Scenario 3: User‑directed changes mid‑plan
While reviewing PR #2, you realize a better design. You tell the agent to change the approach. The agent diverges from the plan, updates the plan artifacts to reflect the new direction, and keeps the series moving without being blocked.

Add it to your agent workflow

Install the skill with a single CLI command:

npx skills add sesori-ai/sesori_apps_monorepo --skill sesori-plan-worker

Once installed, load it in your agent prompt and point it at a plan. For example:
“You are a plan worker. Use the plan at PLAN.md and TRACKER.md to implement the next step. After you open a PR, watch for its merge and then continue.”

The skill’s README explicitly says the user’s instruction is always authoritative—so it won’t refuse a sensible off‑plan request.

Compare before adopting

Before you adopt it, compare against other plan‑execution skills or your own custom prompt. Check the repo’s star count (105 at the time of writing), recent commit activity, and whether the skill’s philosophy matches your own—do you want the agent to auto‑advance after merges, or would you rather manually approve each step? See how it handles plan file updates: does it overwrite or append? And verify it works with your preferred git host and CI triggers.

Why it is worth tracking

The skill has a clear, focused scope and a small but active community (just over 100 stars it’s still early, but it addresses a real pain point). Its behavior of “the plan is an editable guide, not a boundary” is a refreshingly pragmatic approach for agent autonomy. As more teams adopt multi‑step PR workflows, this kind of auto‑advancing worker will become a staple. Keep an eye on its evolution—if you already maintain plans in your repo, it’s worth a trial run today. For more skills in the automation category, see the OpenAgentSkill directory.