executing-plans

Execute written implementation plans in batches with review checkpoints and verification.

1|Updated May 6, 2026
One-click install
npx skills add https://github.com/surfingalien/FinSurfing --skill executing-plans-surfingalien
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: executing-plans
Source: https://github.com/surfingalien/FinSurfing/tree/main/.claude/skills/executing-plans
Command: npx skills add https://github.com/surfingalien/FinSurfing --skill executing-plans-surfingalien

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Implementing a written plan end-to-end without checkpoints risks building on flawed assumptions, skipping verifications, and drifting from the intended design. This Skill structures plan execution into small batches with mandatory review pauses so mistakes are caught early. ## Core Features & Use Cases - Critical Plan Review: Reads the full plan first and raises concerns about vague steps, missing dependencies, or questionable technical choices before any code is written. - Batched Execution with Checkpoints: Executes a default of 3 tasks per batch, runs each task's specified verifications, then reports results and waits for feedback before continuing. - Structured Completion: After all tasks pass, runs the full test suite, confirms a clean working tree, and offers merge, PR, or manual review options. - Use Case: You have a spec from a planning session for a new API endpoint. Invoke this Skill to implement it task by task, pausing after every 3 tasks so a reviewer can course-correct before more work is invested. ## Quick Start Use the executing-plans skill to implement the plan in docs/feature-plan.md, executing three tasks at a time and reporting back for review between batches.

Frequently Asked Questions about executing-plans

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

FAQPage Schema
How do I execute an implementation plan with an AI coding assistant?

Load the plan file, review it critically for gaps or contradictions, then execute tasks in batches of three with verification after each. Report results between batches and wait for feedback before continuing to the next batch.

How many tasks should be executed per batch when implementing a plan?

The default is 3 tasks per batch. Smaller batches keep reporting clear and feedback loops fast; only increase batch size if the plan explicitly says so or your reviewer approves after seeing initial results.

What should I do when a task verification fails during plan execution?

Stop the batch immediately and report the failure instead of marking the task complete. A failed verification signals a flawed plan assumption or misunderstood step, and continuing compounds the error into later dependent tasks.

When should plan execution pause and return to the review step?

Return to review when the plan is updated mid-execution, when a fundamental approach needs rethinking, or when you hit blockers like missing dependencies or unclear instructions. Guessing is slower than asking for clarification.

Should I implement a plan directly on the main branch?

No. Never start implementation on main or master without explicit user consent. Consider using a git worktree or feature branch so work is sandboxed and bad experiments can be abandoned cleanly.