mcp-app-verification
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.
Install / Use
npx skills add a5c-ai/babysitter --skill mcp-app-verificationInstalls into whichever agent you are using.
SKILL.md
Installable skill definition
Quality Score
Category
AutomationSupported Platforms
Our assessment of mcp-app-verification
mcp-app-verification scores 92/100 on our quality scale, 883rd of 2,790 Automation skills we index (top 32%).
Its SKILL.md is 12 KB long, well organised into 57 sections with 5 code examples: a thorough specification that gives an agent plenty to work with.
With 1,809 GitHub stars, it is one of the more widely adopted skills in the catalogue.
Maintenance, license and trust
- The repository was last updated about 2 months ago, so mcp-app-verification is actively maintained.
- Our last check on 2026-10-27 found the source still online.
- It is released under the MIT license, a permissive license that allows use, modification and commercial use with attribution.
- Its trust signals score 100/100, with no cautions. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.
Safety scan
No issues foundOur scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful.
AI review by kimi-k2.7-code on 2026-09-30. Automated pattern scan on 2026-09-30. It catches known dangerous patterns, not every risk — read a skill before letting an agent act on it.
mcp-app-verification compared with similar skills
All 4 of these similar skills score higher than mcp-app-verification; compare them before choosing.
| Skill | Score | Stars | Updated | Format |
|---|---|---|---|---|
| mcp-app-verification (this skill)by a5c-ai | 92 | 1.8k | 2mo ago | SKILL.md |
| Agent-Reachby Panniantong | 100 | 112.3k | 8d ago | CLAUDE.md |
| rufloby ruvnet | 100 | 81.9k | today | CLAUDE.md |
| headroomby headroomlabs-ai | 100 | 79.8k | today | CLAUDE.md |
| CowAgentby zhayujie | 100 | 48.0k | today | CLAUDE.md |
Frequently asked questions
- How do I install mcp-app-verification?
- Run
npx skills add a5c-ai/babysitter --skill mcp-app-verification. The install tabs above show the steps for each supported agent. - Which AI agents does mcp-app-verification work with?
- It is written for Universal, as a SKILL.md file. Other agents that read the same format can often use it too.
- Is mcp-app-verification safe to use?
- Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful. It is MIT-licensed and scores 100/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
- Is mcp-app-verification still maintained?
- The repository was last updated about 2 months ago, so mcp-app-verification is actively maintained.
Skill content
View source on GitHubname: 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]
mcp-app-verification
Run comprehensive verification checklists for MCP Apps covering correctness, compatibility, and migration completeness.
Overview
MCP Apps have several critical invariants that must be verified before deployment. This skill provides systematic verification across multiple dimensions:
- Runtime verification: App loads and functions in basic-host reference
- Pattern verification: Critical code patterns are correct (handler-before-connect, text fallback)
- Build verification: Single-file bundle is valid and complete
- Styling verification: Host theming applies correctly
- CSP verification: All origins declared, no silent failures
- Migration verification: No remaining legacy patterns (OpenAI, old MIME types, snake_case)
Capabilities
basic-host Test Execution
- Build the MCP App
- Start the server
- Launch basic-host reference implementation against the server
- Verify app loads without console errors
- Verify handlers fire correctly
Handler-Before-Connect Invariant
- Search source code for
app.connect()call - Verify ALL handlers (
ontoolinput,ontoolresult,onhostcontextchanged,onteardown) are registered BEFORE connect - Flag violations -- handlers registered after connect will silently not work
Text Fallback Verification
- Search all tool handlers for
contentarray in return value - Verify each tool returns at least one
{ type: 'text', text: '...' }entry - Flag tools that only return
structuredContentwithout text fallback
Resource URI Link Integrity
- Extract all
resourceUrivalues fromregisterAppToolcalls - Extract all URIs from
registerAppResourcecalls - Verify every tool
resourceUrihas a matching registered resource - Flag orphaned resources (registered but not referenced)
Single-File Bundle Verification
- Build the project
- Verify
dist/mcp-app.html(or equivalent) exists - Check the HTML file is self-contained (no external
<script src>,<link href>,<img src>to relative paths) - Verify
vite-plugin-singlefileis in dev dependencies
Host Styling Verification
- Search CSS for
var(--color-*,var(--font-*,var(--border-radius-*patterns - Verify fallback values are present:
var(--color-background-primary, #ffffff)not justvar(--color-background-primary) - Check
onhostcontextchangedhandler exists and applies styling
CSP Verification
- Build and search output for network origins
- Compare against CSP configuration in
registerAppResource - Flag origins present in code but missing from CSP
- Verify conditional origins match between runtime and CSP config
Legacy Pattern Detection (Migration)
- Search for remaining OpenAI patterns:
window.openai.toolInput,window.openai.toolOutput,window.openai - Search for old metadata paths:
openai/ - Search for old MIME types:
text/html+skybridge - Search for hardcoded MIME type:
text/html;profile=mcp-app(should useRESOURCE_MIME_TYPE) - Search for snake_case CSP:
_domains"or_domains:(should be camelCase)
Usage
Full Verification Workflow
# 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 $SERVER_PID
Pattern Verification Commands
# Handler-before-connect check
# Find app.connect() and verify handlers are above it
grep -n 'app\.connect\|\.ontoolinput\|\.ontoolresult\|\.onhostcontextchanged\|\.onteardown' src/main.ts
# Text fallback check
# Every tool handler should return content array
grep -A5 'return {' src/server.ts | grep -c 'content:'
# Resource URI linking
grep 'resourceUri' src/server.ts
grep "registerAppResource" src/server.ts
# RESOURCE_MIME_TYPE usage (not hardcoded)
grep 'RESOURCE_MIME_TYPE' src/server.ts
grep "text/html;profile" src/server.ts # Should NOT match
# CSS variable fallbacks
grep -c 'var(--.*,' src/global.css # Count with fallbacks
grep 'var(--' src/global.css | grep -v ',' # Flag missing fallbacks
Migration Verification (OpenAI -> MCP)
# Server-side legacy patterns
grep -rn 'openai/' src/ # Old metadata paths
grep -rn 'text/html+skybridge' src/ # Old MIME type
grep -rn "text/html;profile=mcp-app" src/ # Hardcoded (use RESOURCE_MIME_TYPE)
grep -rn '_domains"' src/ # Snake_case CSP
grep -rn "_domains:" src/ # Snake_case CSP
# Client-side legacy patterns
grep -rn 'window\.openai\.toolInput' src/
grep -rn 'window\.openai\.toolOutput' src/
grep -rn 'window\.openai' src/
# All should return ZERO matches
Automated Verification Script
#!/bin/bash
# mcp-app-verify.sh - Comprehensive MCP App verification
ERRORS=0
echo "=== MCP App Verification ==="
# 1. Build
echo "[1/8] Building..."
npm run build 2>&1 || { echo "FAIL: Build failed"; ERRORS=$((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
Verification Checklist (Summary)
Core Invariants
- [ ] App builds without errors
- [ ]
dist/mcp-app.htmlexists and is self-contained single file - [ ]
vite-plugin-singlefilein devDependencies - [ ] ALL handlers registered BEFORE
app.connect() - [ ] Every tool returns
content[]with text fallback - [ ] Every tool's
resourceUrimatches a registered resource - [ ]
RESOURCE_MIME_TYPEused (not hardcoded string)
Styling
- [ ] CSS variables use
var(--name, fallback)pattern - [ ]
onhostcontextchangedhandler registered - [ ] App looks correct with host styling (test in basic-host)
- [ ] App looks correct without host styling (standalone fallbacks)
CSP (if applicable)
- [ ] All network origins discovered in build output
- [ ] All origins declared in
resourceDomains/connectDomains/frameDomains - [ ] CSP in correct location:
contents[]ofregisterAppResourceread callback - [ ] Conditional origins: config controls both runtime URL and CSP entry
Migration (if applicable)
- [ ] Zero
window.openaireferences - [ ] Zero
text/html+skybridgereferences - [ ] Zero
openai/metadata paths - [ ] Zero snake_case CSP properties (
_domains) - [ ] Zero hardcoded
text/html;profile=mcp-app(useRESOURCE_MIME_TYPE)
Runtime
- [ ] App loads in basic-host without console errors
- [ ]
ontoolinputfires with tool arguments - [ ]
ontoolresultfires with tool result - [ ] Host styling (theme, fonts, colors) applies correctly
- [ ] Teardown handler fires on app close
Task Definition
const 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`
}
};
}
});
Applicable Processes
- create-mcp-app.js
- add-app-to-mcp-server.js
- convert-web-app-to-mcp.js
- migrate-openai-app-to-mcp.js
External Dependencies
- basic-host reference implementation (from SDK repo clone)
- Build toolchain for the target application
- grep/ripgrep for pattern searching
References
Related Skills
- mcp-tool-resource-pattern
- mcp-host-styling-integration
- mcp-csp-investigation
- single-file-bundling
- mcp-app-scaffolding
Related Agents
- mcp-app-architect
- mcp-ui-developer
- csp-security-auditor
Related Skills
Agent-Reach
112.3kGive your AI agent eyes to see the entire internet. Read & search Twitter, Reddit, YouTube, GitHub, Bilibili, XiaoHongShu — one CLI, zero API fees.
ruflo
81.9k🌊 The original agent harness. Deploy intelligent multi-player swarms, coordinate autonomous workflows, and build conversational AI systems. Features adaptive memory, self-learning intelligence, federation, vector RAG integration, and native Claude Code / Codex / Hermes and many more Integrated
headroom
79.8kCompress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.
CowAgent
48.0kOpen-source personal AI assistant & Agent Harness. Plans tasks, runs tools and skills, self-evolves with memory and knowledge. Multi-agent, multi-model, multi-channel. Lightweight, extensible, one-line install.
Languages
Trust signals
From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.
