hephaestus

Executes planned implementation steps from topic files with strict step-by-step discipline.

Updated Nov 21, 2025
One-click install
npx skills add https://github.com/alatyshau/duet --skill hephaestus-alatyshau
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: hephaestus
Source: https://github.com/alatyshau/duet/tree/main/packages/skills/personas/hephaestus
Command: npx skills add https://github.com/alatyshau/duet --skill hephaestus-alatyshau

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When working through a structured plan, AI assistants often jump ahead, make assumptions, or improvise beyond the plan. This persona enforces disciplined, one-step-at-a-time implementation so code changes stay aligned with the documented plan. ## Core Features & Use Cases - Plan-Driven Execution: Reads the topic file (NARRATIVE, OUTPUTS, PLAN) and executes only the current WIP or next TODO step. - Escalation Protocol: Surfaces conflicts and escalates architecture questions to the Daedalus planning persona instead of guessing. - Clarification Over Assumption: Asks explicit questions when plan steps conflict with outputs or files are missing. - Use Case: During a multi-step feature build, activate this persona to implement exactly one planned step, report what was done, and wait for /done or /next before continuing. ## Quick Start Activate the hephaestus persona and ask it to execute the current step from the topic file's plan.

Frequently Asked Questions about hephaestus

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

FAQPage Schema
How do I make an AI follow a plan step by step?

Use a persona that reads the plan file, finds the current WIP or first TODO step, executes only that step, and waits for an explicit /next or /done command before continuing. This prevents the AI from jumping ahead or improvising.

How to stop AI coding assistants from making assumptions?

Define explicit rules requiring the assistant to ask questions when plan steps conflict with expected outputs or referenced files are missing. The hephaestus persona enforces this with a don't-guess-ask rule and example phrasing.

What is the difference between the Hephaestus and Daedalus personas?

Hephaestus handles implementation and code execution following an existing plan, while Daedalus handles architecture and planning. Hephaestus escalates to Daedalus when a task requires architectural decisions.

When should implementation escalate to architecture planning?

Escalate when a step requires architectural decisions beyond the current plan's scope, or when a conflict appears between plan steps and stated outputs. The persona surfaces conflicts rather than resolving them silently.

What are the limitations of a single-persona implementation approach?

A single persona may blur planning and execution boundaries, leading to scope creep. Separating implementation from architecture personas keeps execution precise but requires explicit escalation handoffs between them.