skills.rest
nrdxp
nrdxpCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

engineering

Enforces root-cause, type-safety, and error-handling rules for code edits and reviews.

10|2|Updated Jan 24, 2026
Search Tags:#type safety#code review#refactoring#error handling#root cause analysis#testing strategy#git hygiene
One-click install
npx skills add https://github.com/nrdxp/predicate --skill engineering-nrdxp
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: engineering
Source: https://github.com/nrdxp/predicate/tree/main/skills/engineering
Command: npx skills add https://github.com/nrdxp/predicate --skill engineering-nrdxp

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? AI-generated code often papers over root causes, swallows errors silently, and drifts from specifications. This Skill anchors every code edit to a standing set of engineering rules so changes stay correct, typed, and reviewable across long sessions. ## Core Features & Use Cases - Always-On Principles: Root-cause-only fixes, no silent failures, boundary validation, strong typing without escape hatches, and mandatory halts on spec/code discrepancies. - By-Moment Reference: Twelve operational rules covering read-first context acquisition, version-aware API stability, concurrent testing strategy, atomic git commits, and plan tracking. - Uncertainty Governance: Numeric uncertainty thresholds (0.0 / 0.1-0.3 / >0.3) that dictate when to proceed, clarify, or halt, plus a failure-recovery protocol. - Use Case: When asked to refactor a Go module, the agent first reads surrounding code and tests, validates inputs at boundaries, writes tests concurrently with the fix, and stops to ask before any breaking API change on a post-1.0 project. ## Quick Start Apply the engineering rules while refactoring this function to handle errors explicitly and add concurrent tests.

Frequently Asked Questions about engineering

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

FAQPage Schema
How do I enforce root-cause fixes instead of workarounds in AI code edits?▼

Apply rules that forbid compatibility shims and symptom-masking fixes, requiring the underlying cause to be corrected. If the fix exceeds local scope, the agent must halt and discuss broader implications before proceeding.

What error handling rules should AI-generated library code follow?▼

Library code must never panic and must return Result or Option types to callers. Every error path is handled or propagated with the causal chain preserved, and messages must state what failed, why, and where.

When should an AI agent stop and ask instead of proceeding with a code change?▼

It should halt when requirements are ambiguous, acceptance criteria are missing, constraints conflict, or uncertainty exceeds the defined threshold. Generating under unvalidated assumptions is explicitly forbidden.

Does this approach handle API breaking changes differently by version?▼

Yes, it checks the project manifest version first. Pre-1.0 projects favor correct design over compatibility, while post-1.0 projects forbid breaking public API changes without explicit user approval.

Why validate inputs only at system boundaries instead of everywhere?▼

Validating at entry points like user input, API responses, and file contents keeps the perimeter strict while trusting internal invariants. Defensive guards inside trusted code obscure where real boundaries are and add noise.

Related Skills

Explore More Software Engineering Skills→
totvstotvs
115

context-map

Generate a context map of files, dependencies, and tests for Protheus ADVPL/TLPP changes.

Official
Intermediate
pandaworks-software-pltpandaworks-software-plt

hermes-commit

Verify code review and lint checks, then execute atomic commits.

Official
Advanced
ArabelaTsoArabelaTso
142

metamorphic-test-generator

Generate test cases by applying metamorphic properties and verifying output relationships.

Community
Intermediate
newkayak12newkayak12
1

architecture-designer

Design high-level system architectures and generate Architecture Decision Records.

Community
Advanced