GSD Executor

Executes GSD plans with atomic commits, deviation handling, and checkpoint protocols.

Updated Mar 11, 2026
One-click install
npx skills add https://github.com/apsulli/gsd-extended --skill gsd-executor-apsulli
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: GSD Executor
Source: https://github.com/apsulli/gsd-extended/tree/main/.agent/skills/executor
Command: npx skills add https://github.com/apsulli/gsd-extended --skill gsd-executor-apsulli

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Executing multi-task development plans with an AI assistant often leads to skipped steps, uncommitted work, and lost state between sessions. This Skill enforces disciplined plan execution with per-task commits, automatic deviation handling, and structured checkpoints so work stays traceable and resumable. ## Core Features & Use Cases - Atomic Task Execution: Executes each task in a PLAN.md file and creates one git commit per task with conventional commit messages. - Automatic Deviation Rules: Auto-fixes bugs, missing critical functionality, and blocking issues, while pausing for user decisions on architectural changes. - Checkpoint & Continuation Protocols: Stops at human-verify, decision, or human-action checkpoints and supports fresh continuation agents that resume from committed state. - Use Case: You have a phase plan with five tasks including a deployment step requiring CLI authentication. The executor completes each task with commits, pauses at the auth gate with exact login instructions, and produces a SUMMARY.md documenting all deviations and verification results. ## Quick Start Ask the AI to execute the current GSD plan file and commit each task atomically while pausing at any checkpoints.

Frequently Asked Questions about GSD Executor

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

FAQPage Schema
How do I execute a GSD plan with an AI assistant?▼

Provide the PLAN.md file to the executor, which loads project state from .gsd/STATE.md, parses tasks and verification criteria, then executes each task sequentially. Each completed task produces an atomic git commit, and the run ends with a SUMMARY.md report.

How does the executor handle work not in the plan?▼

It applies four deviation rules: bugs, missing critical functionality, and blocking issues are fixed automatically and tracked, while architectural changes trigger a checkpoint requiring user decision. All deviations are documented in the final SUMMARY.md.

What happens when a task requires CLI authentication?▼

Authentication errors are treated as expected gates, not failures. The executor stops, returns a human-action checkpoint with exact login steps and a verification command, then a continuation agent resumes after you authenticate.

Can execution resume after a checkpoint pause?▼

Yes. A fresh continuation agent verifies previous task commits exist in git history, skips completed work, and resumes from the specified task based on the checkpoint type, such as approved verification or a selected decision option.

What are the limitations of autonomous plan execution?▼

The executor cannot proceed past checkpoints without user input and requires an initialized .gsd project with STATE.md. Architectural changes like schema modifications or library switches always require explicit user decisions rather than automatic fixes.