skills.rest
HACK-WU
HACK-WUCOMMUNITY
Explore all skills by this author →
GitHub

request-guard

Validates sudden user modification requests against existing conventions before execution.

2|Updated Jun 8, 2026
Search Tags:#design consistency#request validation#ai guardrail#change review#modification check#side effect analysis
One-click install
npx skills add https://github.com/HACK-WU/skills --skill request-guard-hack-wu
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: request-guard
Source: https://github.com/HACK-WU/skills/tree/main/skills/request-guard
Command: npx skills add https://github.com/HACK-WU/skills --skill request-guard-hack-wu

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When users issue short, context-free modification requests ("change this", "add a cache"), AI assistants tend to execute blindly, leading to conflicts with existing design decisions, over-engineering, or unintended side effects. This Skill inserts a lightweight checkpoint before any modification is applied. ## Core Features & Use Cases - Three-Dimension Check: Evaluates each modification request against existing conventions, simpler alternatives, and potential side effects, producing a pass / questionable / blocked verdict in one round. - Decision Memory Lookup: Queries historical decision records via ki-memory-lookup to detect previously rejected approaches and cross-module dependencies before approving changes. - Structured Verdict Output: Produces a formatted report with per-dimension conclusions, reasons, and concrete alternative suggestions when a request is questioned or blocked. - Use Case: A user suddenly says "convert all synchronous requests to async". The Skill checks the design documents, finds some interfaces must remain synchronous, flags the over-broad scope, and blocks the change with a recommendation to run deeper requirement analysis instead. ## Quick Start Before applying my requested change, run a request-guard check on it and tell me whether it passes, is questionable, or should be blocked.

Frequently Asked Questions about request-guard

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

FAQPage Schema
How do I stop an AI from blindly executing sudden code change requests?▼

Insert a pre-execution checkpoint that evaluates each modification request against existing design conventions, alternative solutions, and side effects. The request-guard Skill outputs a pass, questionable, or blocked verdict with reasons before any change is applied.

What checks should run before applying a user's modification request?▼

Three dimensions: whether the request conflicts with existing design documents or code conventions, whether a simpler or more fundamental alternative exists, and whether the change introduces side effects on other modules. Each dimension yields a pass, warning, or block result.

When should modification request validation be skipped?▼

Skip the check when the user has discussed the change over multiple rounds with full context, when the user explicitly confirms the approach, or when the change is trivial such as a typo or formatting fix. Over-checking minor edits is an anti-pattern.

What is the difference between request-guard and requirement-mining?▼

Request-guard is a lightweight one-round check run before executing a modification, while requirement-mining performs deep requirement analysis. When request-guard detects serious problems, it recommends escalating to requirement-mining rather than analyzing deeply itself.

Can a blocked modification request still be executed?▼

Yes. The final decision always belongs to the user. Even when the check returns a blocked verdict with reasons and alternatives, the change is executed if the user insists, and the Skill does not repeatedly block the same request.

Related Skills

Explore More Software Engineering Skills→
opensage-agentopensage-agent
113

grill-with-docs

Validate design plans against domain models and architectural decision records.

Official
Advanced
jianshuojianshuo
114

wjs-evaling-voicedrop-prompts

Evaluate VoiceDrop mining prompt changes with local evaluations and pairwise judgments.

Community
Advanced
pagerguildpagerguild

code-review-pipeline

Coordinate staged multi-agent code review across pre-commit and pull-request workflows.

Official
Advanced
LayerDynamicsLayerDynamics

diff-review

Review current git diff changes for bugs, security issues, and style inconsistencies.

Community
Intermediate