engineering-loop

Routes engineering tasks between direct edits, implementation, design interrogation, and structural review workflows.

Updated Sep 17, 2026
One-click install
npx skills add https://github.com/jordizle/claude-skills --skill engineering-loop-jordizle
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: engineering-loop
Source: https://github.com/jordizle/claude-skills/tree/main/engineering-loop
Command: npx skills add https://github.com/jordizle/claude-skills --skill engineering-loop-jordizle

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Engineering tasks vary wildly in scope, yet teams often apply the same heavyweight process to everything or skip process entirely. This Skill acts as a coordinator that picks the smallest appropriate workflow for each task, from a direct one-line edit to a full design-interrogation-plus-review pipeline, and re-routes mid-flight when new evidence emerges. ## Core Features & Use Cases - Adaptive routing: Chooses between direct implementation, poteto-mode execution, grill-with-docs decision work, and thermo-nuclear structural review based on the nature of the task, not on which skills happen to be installed. - ADR lifecycle management: Discovers only relevant ADRs, interprets proposed/accepted/superseded statuses correctly, and keeps the ADR index consistent whenever decisions change. - Evidence-driven re-routing: When implementation contradicts an accepted or proposed ADR, it pauses the affected branch, surfaces the evidence, and routes back through the decision layer instead of silently violating architecture. - Use Case: You ask for an ambiguous new feature. The Skill hydrates repository context, invokes grill-with-docs to resolve the design and draft a proposed ADR, hands implementation to poteto-mode, verifies correctness, runs thermo-nuclear review, and finally accepts the ADR with an updated index. ## Quick Start Ask the assistant to implement a bug fix or feature using the engineering-loop workflow and let it choose the appropriate route automatically.

Frequently Asked Questions about engineering-loop

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

FAQPage Schema
How do I choose the right workflow for a coding task?

The engineering-loop skill classifies the task after hydrating repository context and picks the smallest safe workflow. Small obvious changes go direct, defined work goes to poteto-mode, unresolved consequential decisions go to grill-with-docs first, and structurally risky changes get thermo-nuclear review after correctness verification.

When should I use grill-with-docs versus direct implementation?

Use grill-with-docs only when a consequential ambiguity or architectural decision genuinely blocks implementation, such as conflicting interpretations of behavior or a clash with an accepted ADR. Task size alone is not a reason; if an accepted ADR already resolves the question, proceed directly to implementation.

Does engineering-loop require an ADR index to work?

No, an ADR index is optional. The skill uses docs/adr/README.md as a discovery aid when present, but individual ADRs remain the authoritative source. It will not create an index unless the user requests one or repository conventions require it.

What happens when implementation contradicts an accepted ADR?

The skill pauses the affected branch, surfaces the documented decision, the conflicting evidence, and the consequences, then routes through grill-with-docs to update the decision. It never silently violates, reinterprets, or works around an accepted ADR.

When is thermo-nuclear code review not needed?

Thermo-nuclear review is skipped for small, obvious, local changes with negligible structural risk. It is warranted only when changes introduce new abstractions, alter module boundaries, add significant branching or state, or otherwise carry meaningful maintainability risk, and always after correctness verification.