Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add varunk130/ai-ux-skill-library --skill ai-safety-guardrailsgit clone --depth 1 https://github.com/varunk130/ai-ux-skill-libraryWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/varunk130/ai-ux-skill-library/ai-safety-guardrails)<a href="https://agentmods.dev/skills/varunk130/ai-ux-skill-library/ai-safety-guardrails"><img src="https://agentmods.dev/badge/skills/varunk130/ai-ux-skill-library/ai-safety-guardrails/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/varunk130/ai-ux-skill-library/ai-safety-guardrails"><img src="https://agentmods.dev/badge/skills/varunk130/ai-ux-skill-library/ai-safety-guardrails.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00063 | $0.02076 |
| Opus 5 | $0.00032 | $0.01038 |
| Sonnet 5 | $0.00013 | $0.00415 |
| Haiku 4.5 | $0.00006 | $0.00208 |
Grade A, and why
ai-safety-guardrails scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 12d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 173 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AI Safety Guardrails
Design the user-facing layer of AI safety - how products prevent harm, communicate restrictions, handle sensitive content, and maintain accountability. The SHIELD framework ensures safety mechanisms protect users without destroying the user experience.
Core Principle
Safety and usability are not opposites. The best safety design is invisible when everything is fine and clear when it matters. Safety should feel like a guardrail on a mountain road - you forget it's there until you need it, and when you need it, you're grateful it's solid.
The SHIELD Framework
| Letter | Principle | Design Question |
|---|---|---|
| S | Scope the Risks | Have you mapped every way this AI could produce harm? |
| H | Human Oversight Gates | Are there human checkpoints before high-risk AI actions take effect? |
| I | Inform on Restriction | When the AI restricts output, does it explain why and what alternatives exist? |
| E | Escalation Paths | Can users challenge a restriction through a clear, fair process? |
| L | Log Everything | Is there an auditable record of safety decisions? |
| D | Degrade Gracefully | When safety mechanisms activate, does the product still provide value? |
The Harm Taxonomy for AI Products
Design different safety responses for different harm types:
| Harm Category | Examples | Severity | UX Response |
|---|---|---|---|
| Misinformation | Hallucinated facts, fabricated citations, false statistics | High | Citation requirements, confidence indicators, verification prompts |
| Bias amplification | Stereotyping in outputs, skewed recommendations, unfair treatment | High | Bias detection indicators, diverse output prompts, fairness disclaimers |
| Privacy leakage | AI revealing personal data, memorized training content | Critical | Input sanitization, output scanning, PII detection alerts |
| Harmful instructions | Dangerous activities, self-harm, illegal guidance | Critical | Hard block + resource referral (crisis lines, safety info) |
| Manipulation | Persuasion without disclosure, dark patterns, emotional exploitation | High | Transparency requirements, intent disclosure |
| Overreliance | Users making critical decisions solely based on AI output | Medium | Decision-support framing, "consult a professional" nudges |
| Exclusion | AI that works poorly for certain languages, cultures, or accessibility needs | Medium | Coverage transparency, alternative pathways, inclusion testing |
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 12d ago First seen · 173 lines · 63 tokens per session scan A 1a3768becfd7
ai-safety-guardrails is a skill published in the GitHub repository varunk130/ai-ux-skill-library (3 stars, last pushed 1mo ago), licensed MIT. It adds 63 tokens to every session and 2,076 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
design-dna
Extract, define, and apply design DNA across three dimensions: design system (tokens), design style (qualitative feel), and visual effects (Canvas, WebGL, 3D, particles, shaders, scroll effects, etc.). Use this skill when: (1) a user wants to see the full 3-dimension design structure/schema, (2) a user provides…
ss-dial
Turn ONE design axis up or down as a coordinated, deterministic transform — "denser", "sharper corners", "more muted", "bolder", "flatter", "livelier". Not a vibe the model reinterprets each time; a defined ramp that moves many tokens together, respects the guardrails (8px grid, a11y floors, single accent…
ss-reference
Compile screenshots, URLs, Figma exports, or an existing UI into a project-local StyleSeed output grammar with evidence, tokens, confidence, anti-patterns, and a validation screen. Use when the user supplies a design reference that StyleSeed does not already model.
bim-classification-ai
Classify BIM elements using AI and standard classification systems. Map elements to UniFormat, MasterFormat, OmniClass, and CWICR codes.
032-architecture-adr-non-functional-requirements
Facilitates conversational discovery to create Architectural Decision Records (ADRs) for non-functional requirements using the ISO/IEC 25010:2023 quality model. Use when the user wants to document quality attributes, NFR decisions, security/performance/scalability architecture, or design systems with measurable…
component-family-consistency
Buttons, inputs, pills, badges, calendars, and other interactive components form a visual family — they share the same border-radius, colour logic, shadow scale, border style, and spacing rhythm. Inconsistency between them breaks the sense of a coherent product. Use when building or reviewing a component library…