Token导航 LogoToken导航TokenDH.com
研究检索操作浏览器github未标认证来源可访问许可证需确认审计提醒

deer-sense鹿感

Agent Skill

deer-sense 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

总安装

1,591

周安装

65

GitHub Stars

4

下载量

510
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

unknown

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:deer-sense(鹿感)
来源仓库:https://github.com/autumnsgrove/groveengine
仓库路径:skills/deer-sense
安装命令:
npx skills add https://github.com/autumnsgrove/groveengine --skill deer-sense
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/autumnsgrove/groveengine --skill deer-sense

简介

Deer Sense 专注数字界面无障碍检测,模拟残障用户交互体验。

  • 自动识别色彩对比度、键盘导航与屏幕阅读器兼容性问题。
  • 适用于新建组件审核与既有系统改造前的 accessibility 评估。
  • 输出具体修复建议而非泛泛而谈,直接关联 WAI-ARIA 标准条目。
  • deer-sense 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

SKILL.md

Deer Sense

The deer moves through the forest with heightened awareness. It hears what others miss—a twig snapping underfoot, a bird's warning call. It notices what blocks the path. The deer guides the herd around danger, ensuring everyone can travel safely. In the digital forest, the deer senses barriers that stop some wanderers from finding their way.

When to Activate

  • User asks to "check accessibility" or "a11y audit"
  • User says "make this accessible" or "screen reader test"
  • User calls /deer-sense or mentions deer/accessibility
  • Building new UI components
  • Reviewing existing interfaces
  • After visual redesigns
  • Before major releases
  • When accessibility complaints arise

Pair with: chameleon-adapt for accessible UI implementation


The Sense

LISTEN --> SCAN --> TEST --> GUIDE --> PROTECT
   |         |        |        |          |
 Hear     Look for  Validate  Fix      Prevent
 Needs    Barriers   Paths   Issues   Regression

Phase 1: LISTEN

*The deer's ears twitch, hearing what others miss before the danger arrives...*

Understand who we're building for and what barriers they face. Map disability types to assistive technologies. Establish WCAG level targets for this project.

  • Identify relevant disability types: visual, motor, cognitive, auditory — and which assistive technologies they use
  • Confirm WCAG target: Grove standard is WCAG 2.1 AA minimum
  • Map common barriers for this specific interface: color contrast, missing alt text, keyboard traps, small touch targets, missing form labels, confusing navigation
  • Note Grove-specific considerations: GroveTerm components, glass surfaces, seasonal animations, reduced motion

Reference: Load references/automated-scanning.md for the full disability/assistive technology table and WCAG level guide, and references/grove-a11y-patterns.md for Grove-specific barrier patterns


Phase 2: SCAN

*The deer's eyes scan the forest floor, spotting what blocks the path before setting a hoof forward...*

Run automated scanning to surface mechanical violations quickly. Automated tools catch approximately 30% of real issues — use them to clear obvious problems before manual testing begins.

  • Install and run axe-core CLI: npx axe https://localhost:5173 --tags wcag2aa
  • Review Svelte compiler's built-in a11y warnings (missing alt, unlabeled inputs, div with click)
  • Check the common issues checklist: images, forms, navigation, color contrast, page structure, language attribute
  • Note violations and group by severity: blockers (missing labels, keyboard traps) vs. warnings (suboptimal alt text)

Component isolation audit with Showroom:

Before scanning full pages, audit individual components in isolation using Showroom. Showroom checks focus styles, heading hierarchy, and interactive element compliance at the component level — catching a11y issues that get masked by page context.

# Audit each component the deer is reviewing
uv run --project tools/glimpse glimpse showroom \
  libs/engine/src/lib/ui/components/[path-to-component].svelte

# The audit checks: focus styles on interactive elements, heading hierarchy,
# image alt text, color contrast via computed styles, spacing grid compliance

This is a required gate for component-level a11y work. A component that lacks focus indicators or has broken heading hierarchy must be caught here, not deferred to page-level scanning.

Visual scan with Glimpse:

Capture the page to see what the rendered result actually looks like. Console logs often reveal hidden a11y issues:

# Prerequisite: seed the database if not already done
uv run --project tools/glimpse glimpse seed --yes

# Capture with console logs — errors often reveal missing ARIA, broken refs
# Local routing uses ?subdomain= for tenant isolation; --auto starts the dev server
uv run --project tools/glimpse glimpse capture \
  "http://localhost:5173/[page]?subdomain=midnight-bloom" \
  --season autumn --theme dark --logs --auto

# Check dark mode contrast — capture both themes side by side
uv run --project tools/glimpse glimpse matrix \
  "http://localhost:5173/[page]?subdomain=midnight-bloom" \
  --themes light,dark --logs --auto

# Browse interactively — does tab order make visual sense?
uv run --project tools/glimpse glimpse browse \
  "http://localhost:5173/[page]?subdomain=midnight-bloom" \
  --do "click first interactive element, press Tab, press Tab, press Tab" \
  --screenshot-each --auto

Review the screenshots: Are focus indicators visible? Do glass surfaces maintain contrast? Are touch targets visually large enough?

Reference: Load references/automated-scanning.md for axe-core commands, Lighthouse CI setup, ESLint plugin config, CI integration YAML, and the full issues checklist


Phase 3: TEST

*The deer tests each path personally, because knowing the map is not the same as walking the trail...*

Manual testing catches what automation misses. Test keyboard navigation, screen reader output, reduced motion, and zoom. Work through the test scenarios relevant to this interface.

  • Keyboard testing: Tab order, Enter/Space activation, focus indicators, escape key, skip links, no traps
  • Screen reader testing (VoiceOver or NVDA): page title, heading navigation, landmarks, images, buttons, forms, live regions
  • Zoom testing at 200%: no horizontal scroll, all content visible, no overlapping elements
  • Reduced motion: verify all animations are conditional on prefers-reduced-motion

Reference: Load references/keyboard-testing.md for the keyboard checklist, focus trap patterns, modal testing, touch target requirements, and conditional field test scenarios. Load references/screen-reader-testing.md for screen reader tools, ARIA patterns, semantic HTML fixes, and the ARIA checklist.


Phase 4: GUIDE

*The deer guides the herd around the fallen tree, showing the clear path with a gentle turn of its head...*

Fix accessibility issues with proper implementations. Prefer semantic HTML over ARIA when possible. Every fix should leave the interface clearer for everyone.

  • Fix color contrast issues: use Grove palette tokens which are pre-tested for contrast ratios
  • Replace div soup with semantic HTML (article, button, nav, main)
  • Add focus management: openModal() focuses first element, closeModal() returns focus to trigger
  • Add ARIA labels to icon-only buttons; add live regions for dynamic status updates
  • Fix form accessibility: for/id pairs, aria-describedby for errors, role="alert" on error messages
  • Ensure 44px minimum touch targets on all interactive Grove elements

Reference: Load references/screen-reader-testing.md for ARIA label patterns, semantic HTML fixes, and form accessibility code. Load references/grove-a11y-patterns.md for touch target patterns, glass card focus styles, GroveTerm requirements, and reduced motion store usage.


Phase 5: PROTECT

*The deer stands watch, ensuring the path stays clear long after this audit is done...*

Prevent future accessibility regressions. Add automated checks to CI and maintain documentation so the forest stays walkable for every wanderer.

  • Add axe-core to vitest/puppeteer test suite: any violation fails the test
  • Add eslint-plugin-svelte-a11y for development-time warnings
  • Configure CI to run axe-core on every pull request
  • Update accessibility documentation with standards, testing process, and tooling
  • Generate the final audit report: pages tested, automated results, manual testing results, issues fixed, ongoing protections

Reference: Load references/automated-scanning.md for the CI YAML, ESLint plugin setup, axe-core in vitest, and the documentation template. Load references/screen-reader-testing.md for the final audit report template.


Reference Routing Table

PhaseReferenceLoad When
LISTENreferences/automated-scanning.mdAlways (disability types, WCAG levels)
LISTEN (Grove)references/grove-a11y-patterns.mdAny Grove component audit
SCANreferences/automated-scanning.mdAlways (axe-core, Lighthouse, checklist)
TEST (keyboard)references/keyboard-testing.mdManual keyboard and zoom testing
TEST (screen reader)references/screen-reader-testing.mdScreen reader testing scenarios
GUIDEreferences/screen-reader-testing.md + references/grove-a11y-patterns.mdFixing issues
PROTECTreferences/automated-scanning.md + references/screen-reader-testing.mdCI setup, final report

Deer Rules

Awareness

Notice what others miss. Accessibility barriers hide in details — the focus ring that's technically present but invisible over a glass surface, the touch target that's 42px instead of 44.

Inclusion

Design for everyone. The forest is for all wanderers. Accessibility benefits everyone: keyboard shortcuts help power users, captions help people in noisy environments, clear error messages help everyone.

Persistence

Accessibility is ongoing, not one-time. Every new component is an opportunity to get it right from the start, and every PR without a11y checks is a chance to let a barrier back in.

Communication

Use sensory metaphors:

  • "Listening for barriers..." (understanding needs)
  • "Scanning the path..." (automated testing)
  • "Testing the route..." (manual verification)
  • "Guiding around obstacles..." (fixing issues)
  • "The path is clear." (audit complete)

Anti-Patterns

The deer does NOT:

  • Skip testing with real assistive technology
  • Rely only on automated tools (they catch ~30% of real issues)
  • Use "accessibility overlays" or third-party widget patches (they don't work)
  • Ignore cognitive accessibility — confusing navigation is a barrier too
  • Treat a11y as an afterthought added after "the design is done"
  • Suppress focus outlines without providing visible alternatives
  • Use color alone to convey meaning

Example Audit

User: "Audit the new dashboard for accessibility"

Deer flow:

  1. LISTEN — "Dashboard has complex data, charts, tables. Users: screen reader, keyboard, motor. WCAG AA target. Grove glass cards with custom focus needed."
  2. SCAN — "axe-core: 5 violations. Missing alt on charts, low contrast on metrics, no table headers. Svelte warns: 2 div with onclick."
  3. TEST — "Keyboard: Can't reach filter dropdowns (trap in date picker). Screen reader: Tables not navigable (no scope). Zoom: Sidebar overlaps main content at 200%."
  4. GUIDE — "Add aria-labels to charts, increase metric contrast to 4.5:1 using Grove tokens, add proper table markup with <th scope>, make sidebar responsive, replace date picker with keyboard-accessible version."
  5. PROTECT — "axe-core added to CI, eslint-plugin-svelte-a11y enabled, audit report written, accessibility section added to component docs."

Quick Decision Guide

SituationAction
New componentSemantic HTML first, test keyboard/screen reader before shipping
Color choiceCheck contrast ratio (4.5:1 minimum — use Grove palette tokens)
Interactive elementKeyboard accessible, visible focus-visible style required
ImageDescriptive alt text (or alt="" if purely decorative)
FormAssociate labels, error messages with aria-describedby, required indicators
AnimationConditional on $reducedMotion store — no exceptions
GroveTerm/GroveTextLoad references/grove-a11y-patterns.md for required ARIA attributes
Touch target44px minimum — use padding if visual size must stay smaller

Integration with Other Skills

Before Sensing:

  • chameleon-adapt — If building new UI, design with accessibility in mind from the start

After Sensing:

  • chameleon-adapt — For accessible UI implementation of fixes
  • beaver-build — Add accessibility regression tests after fixing issues
  • owl-archive — Document accessibility standards for the project

*The forest welcomes all who seek it. Remove the barriers.*

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

Codex

39.06%
按下载量换算199

Claude

27.8%
按下载量换算142

Cursor

16.95%
按下载量换算86

Gemini CLI

8.79%
按下载量换算45

安全审计

Gen Agent Trust Hub

通过

Socket

可疑

Snyk

通过

权限和风险

操作浏览器

该 Skill 可能涉及浏览器控制能力,使用时可能读取或操作网页内容,需要在受控环境中确认权限边界。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。来源安全扫描存在 warning/failed 结果,不能写成本站确认安全。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills