skills.rest
microsoft
microsoftOFFICIAL
Explore all skills by this author →
Website𝕏@OpenAtMicrosoftGitHub

prompt-analyze

Reviews prompt artifacts read-only by routing static and behavior analysis to hve-builder review mode.

1.4k|284|Updated Nov 2, 2025
Search Tags:#static analysis#read-only#Prompt Engineering#prompt review#behavior testing#gitHub copilot#hve-builder
One-click install
npx skills add https://github.com/microsoft/hve-core --skill prompt-analyze
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: prompt-analyze
Source: https://github.com/microsoft/hve-core/tree/main/.github/skills/hve-core/prompt-analyze
Command: npx skills add https://github.com/microsoft/hve-core --skill prompt-analyze

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Teams maintaining prompt-engineering artifacts (prompts, instructions, agents, skills) need a consistent, read-only way to evaluate quality without accidentally modifying source files or relying on retired analysis workflows.

Core Features & Use Cases

  • Read-only artifact review: Analyzes prompt, instruction, agent, subagent, skill, reference, and template files without altering them.
  • Static plus behavior analysis: Routes targets through hve-builder review mode, producing a static verdict and behavior-test results with configurable fidelity (simulation or native).
  • Structured findings: Returns severity-graded findings, durable report links, and an overall outcome (Pass, Revise, Deferred, or Blocked).
  • Use Case: A developer updates a custom Copilot prompt file and wants an independent quality review with behavior testing before committing, while keeping the source file untouched.

Quick Start

Ask the assistant to run prompt-analyze on your prompt file, for example: review my prompt artifact at .github/prompts/deploy.prompt.md and report the static and behavior verdicts.

Frequently Asked Questions about prompt-analyze

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

FAQPage Schema
How do I review a GitHub Copilot prompt file without modifying it?▼

Invoke prompt-analyze with the promptFiles parameter pointing to your prompt, instruction, or agent file. It routes the review through hve-builder in read-only review mode, so source artifacts remain unchanged while you receive static and behavior verdicts.

What file types can prompt-analyze review?▼

It reviews prompt, instruction, agent, subagent, skill, reference, and template files. You pass them via the promptFiles input, or it infers currently open or attached prompt-engineering artifacts when omitted.

What is the difference between simulation and native behavior-test fidelity?▼

Simulation fidelity models behavior without real runtime execution, while native fidelity runs actual runtime tests subject to HVE Builder Tester safety preconditions. Simulation results are never reported as native runtime evidence.

Can prompt-analyze fix or refactor my prompt files?▼

No. It is strictly read-only and cannot research, fix, refactor, or create source artifacts. For approved improvements it hands off to prompt-builder, and for behavior-preserving cleanup it recommends prompt-refactor.

What do the Pass, Revise, Deferred, and Blocked outcomes mean?▼

Pass means the review completed successfully. Revise indicates findings requiring changes, Deferred means behavior testing was postponed with a stated reason, and Blocked means a precondition failed. Each non-Pass outcome includes a rerun condition.

Related Skills

Explore More Software Engineering Skills→
dhar174dhar174
1

tldr-prompt

Generate tldr summaries for GitHub Copilot customization files and documentation URLs.

Community
Intermediate
korwlik23korwlik23

code-review-refactor

Review code and refactor guidance with evidence-backed, severity-classified findings.

Community
Advanced
testdoubletestdouble
198

refactor

Refactor code iteratively with test-gated validation after each step.

Official
Advanced
TenureAITenureAI
51

experiment-execution

Execute AI/ML experiments locally or remotely with explicit environment and runtime controls.

Official
Advanced