菜单
Comprehensive verification checklists for MCP Apps. Tests with basic-host reference, validates handler-before-connect, text fallback, resource URI linking, single-file bundling, host styling, CSP, and legacy pattern detection.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/a5c-ai/babysitter --skill mcp-app-verification命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
正在显示 SKILL.md
基于 SOC 职业分类
| name | mcp-app-verification |
| description | Comprehensive verification checklists for MCP Apps. Tests with basic-host reference, validates handler-before-connect, text fallback, resource URI linking, single-file bundling, host styling, CSP, and legacy pattern detection. |
| allowed-tools | Read, Bash, Glob, Grep |
| graph | {"domains":["domain:software-engineering"],"specializations":["specialization:ai-agents-conversational"],"skillAreas":["skill-area:mcp-server-implementation","skill-area:agent-simulation-testing"],"roles":["role:backend-engineer","role:fullstack-engineer"],"workflows":["workflow:feature-development"],"topics":["topic:api-design"]} |
Run comprehensive verification checklists for MCP Apps covering correctness, compatibility, and migration completeness.
MCP Apps have several critical invariants that must be verified before deployment. This skill provides systematic verification across multiple dimensions:
app.connect() callontoolinput, ontoolresult, onhostcontextchanged, onteardown) are registered BEFORE connectcontent array in return value{ type: 'text', text: '...' } entrystructuredContent without text fallbackresourceUri values from registerAppTool callsregisterAppResource callsresourceUri has a matching registered resourcedist/mcp-app.html (or equivalent) exists<script src>, <link href>, <img src> to relative paths)vite-plugin-singlefile is in dev dependenciesvar(--color-*, var(--font-*, var(--border-radius-* patternsvar(--color-background-primary, #ffffff) not just var(--color-background-primary)onhostcontextchanged handler exists and applies stylingregisterAppResourcewindow.openai.toolInput, window.openai.toolOutput, window.openaiopenai/text/html+skybridgetext/html;profile=mcp-app (should use RESOURCE_MIME_TYPE)_domains" or _domains: (should be camelCase)# Step 1: Build the project
npm run build
# Step 2: Verify single-file bundle
ls -la dist/mcp-app.html
# Should be a single file with all assets inlined
# Step 3: Check for external references in bundle
grep -E '<script src="|<link.*href="|<img src="(?!data:)' dist/mcp-app.html
# Should return NOTHING (all assets inlined)
# Step 4: Start server
npm run serve &
SERVER_PID=$!
# Step 5: Test with basic-host
cd /tmp/mcp-ext-apps/examples/basic-host
SERVERS='["http://localhost:3001/mcp"]' npm run start
# Verify: app loads, handlers fire, styling applies
# Step 6: Stop server
kill $SERVER_PID
# Handler-before-connect check
# Find app.connect() and verify handlers are above it
grep -n 'app\.connect\|\.ontoolinput\|\.ontoolresult\|\.onhostcontextchanged\|\.onteardown' src/main.ts
# Text fallback check
# Every tool handler should return content array
grep -A5 'return {' src/server.ts | grep -c 'content:'
# Resource URI linking
grep 'resourceUri' src/server.ts
grep "registerAppResource" src/server.ts
# RESOURCE_MIME_TYPE usage (not hardcoded)
grep 'RESOURCE_MIME_TYPE' src/server.ts
grep "text/html;profile" src/server.ts # Should NOT match
# CSS variable fallbacks
grep -c 'var(--.*,' src/global.css # Count with fallbacks
grep 'var(--' src/global.css | grep -v ',' # Flag missing fallbacks
# Server-side legacy patterns
grep -rn 'openai/' src/ # Old metadata paths
grep -rn 'text/html+skybridge' src/ # Old MIME type
grep -rn "text/html;profile=mcp-app" src/ # Hardcoded (use RESOURCE_MIME_TYPE)
grep -rn '_domains"' src/ # Snake_case CSP
grep -rn "_domains:" src/ # Snake_case CSP
# Client-side legacy patterns
grep -rn 'window\.openai\.toolInput' src/
grep -rn 'window\.openai\.toolOutput' src/
grep -rn 'window\.openai' src/
# All should return ZERO matches
#!/bin/bash
# mcp-app-verify.sh - Comprehensive MCP App verification
ERRORS=0
echo "=== MCP App Verification ==="
# 1. Build
echo "[1/8] Building..."
npm run build 2>&1 || { echo "FAIL: Build failed"; ERRORS=$((ERRORS+1)); }
# 2. Single-file bundle
echo "[2/8] Checking single-file bundle..."
if [ ! -f dist/mcp-app.html ]; then
echo "FAIL: dist/mcp-app.html not found"
ERRORS=$((ERRORS+1))
fi
# 3. No external references
echo "[3/8] Checking for external references..."
EXT_REFS=$(grep -cE 'src="(?!data:)[^"]+"|href="(?!data:)[^"]+\.css"' dist/mcp-app.html 2>/dev/null || echo "0")
if [ "$EXT_REFS" -gt 0 ]; then
echo "WARN: Found $EXT_REFS potential external references"
fi
# 4. Handler-before-connect
echo "[4/8] Checking handler-before-connect..."
CONNECT_LINE=$(grep -n 'app\.connect()' src/main.ts* 2>/dev/null | head -1 | cut -d: -f2)
if [ -n "$CONNECT_LINE" ]; then
LATE_HANDLERS=$(grep -n '\.on\(toolinput\|toolresult\|hostcontextchanged\|teardown\)' src/main.ts* 2>/dev/null | awk -F: -v cl="$CONNECT_LINE" '$2 > cl')
if [ -n "$LATE_HANDLERS" ]; then
echo "FAIL: Handlers registered after app.connect()"
ERRORS=$((ERRORS+1))
fi
fi
# 5. Text fallback
echo "[5/8] Checking text fallback..."
# (manual review needed for complex cases)
# 6. RESOURCE_MIME_TYPE
echo "[6/8] Checking RESOURCE_MIME_TYPE usage..."
HARDCODED=$(grep -rn "text/html;profile=mcp-app" src/ 2>/dev/null | wc -l)
if [ "$HARDCODED" -gt 0 ]; then
echo "FAIL: Hardcoded MIME type found (use RESOURCE_MIME_TYPE)"
ERRORS=$((ERRORS+1))
fi
# 7. CSS fallbacks
echo "[7/8] Checking CSS variable fallbacks..."
NO_FALLBACK=$(grep 'var(--' src/*.css 2>/dev/null | grep -v ',' | wc -l)
if [ "$NO_FALLBACK" -gt 0 ]; then
echo "WARN: $NO_FALLBACK CSS variables without fallback values"
fi
# 8. Legacy patterns (migration)
echo "[8/8] Checking for legacy patterns..."
LEGACY=$(grep -rn 'window\.openai\|text/html+skybridge\|_domains"' src/ 2>/dev/null | wc -l)
if [ "$LEGACY" -gt 0 ]; then
echo "FAIL: $LEGACY legacy patterns found"
ERRORS=$((ERRORS+1))
fi
echo ""
if [ "$ERRORS" -eq 0 ]; then
echo "PASS: All verification checks passed"
else
echo "FAIL: $ERRORS verification errors found"
fi
exit $ERRORS
dist/mcp-app.html exists and is self-contained single filevite-plugin-singlefile in devDependenciesapp.connect()content[] with text fallbackresourceUri matches a registered resourceRESOURCE_MIME_TYPE used (not hardcoded string)var(--name, fallback) patternonhostcontextchanged handler registeredresourceDomains / connectDomains / frameDomainscontents[] of registerAppResource read callbackwindow.openai referencestext/html+skybridge referencesopenai/ metadata paths_domains)text/html;profile=mcp-app (use RESOURCE_MIME_TYPE)ontoolinput fires with tool argumentsontoolresult fires with tool resultconst mcpAppVerificationTask = defineTask({
name: 'mcp-app-verification',
description: 'Run comprehensive MCP App verification',
inputs: {
projectDir: { type: 'string', required: true },
isMigration: { type: 'boolean', default: false },
migrationSource: { type: 'string', default: '' },
checkCsp: { type: 'boolean', default: true }
},
outputs: {
passed: { type: 'boolean' },
errors: { type: 'array' },
warnings: { type: 'array' },
artifacts: { type: 'array' }
},
async run(inputs, taskCtx) {
return {
kind: 'skill',
title: `Verify MCP App: ${inputs.projectDir}`,
skill: {
name: 'mcp-app-verification',
context: {
projectDir: inputs.projectDir,
isMigration: inputs.isMigration,
migrationSource: inputs.migrationSource,
checkCsp: inputs.checkCsp,
instructions: [
'Build the application',
'Verify single-file bundle integrity',
'Check handler-before-connect invariant',
'Verify text fallback in all tools',
'Validate resource URI linking',
'Check CSS variable fallbacks',
inputs.checkCsp ? 'Verify CSP completeness' : null,
inputs.isMigration ? 'Search for legacy patterns' : null,
'Test with basic-host reference'
].filter(Boolean)
}
},
io: {
inputJsonPath: `tasks/${taskCtx.effectId}/input.json`,
outputJsonPath: `tasks/${taskCtx.effectId}/result.json`
}
};
}
});
Reference for querying the Atlas knowledge graph through its MCP tools — the SECONDARY enrichment/comparison layer that adds best-practice context to systems you have ALREADY scanned from your real sources (`az`, repos, dirs). Use when you need to look up nodes, edges, kinds, clusters, stats, or wiki pages in Atlas to compare against your real inventory. (atlas graph, query atlas, atlas mcp, search the graph, graph neighbors, atlas record, atlas kinds, enrichment layer)
Atlas turns your STATED NEED into a real systems atlas by SCANNING your actual sources (Azure via `az`, git repos, local dirs) and process/data mining them, THEN enriching against the Atlas knowledge graph. Use this skill when asked to inventory/map your real systems, scan your cloud + repos + directories, mine the real processes or data they contain, or collect their real constraints/gotchas. (atlas, scan my systems, inventory our azure account, map my repos, real systems atlas, process mining, data mining, collect nuances, system discovery)
This skill should be used when the user asks to "find skills in the wild", "assimilate popular workflows", "discover SKILL.md files in repos", "research external skills", "find workflow patterns", "survey the skill landscape", "what skills exist out there", or wants to investigate public repositories for extractable processes, babysitter plugins, and reusable procedural insights. Searches GitHub for SKILL.md files, classifies repos by archetype, and maintains structured research under docs/reference-repos/.