Coding agent files, skills and runtime checks

Give AI agents better boundaries.

Review the instructions, skills, configuration, authentication, and distribution metadata that surround coding agents. Each result explains the issue instead of returning an opaque score.

AI agent tool collection

Stateless runtime

MCP 2026 Stateless Check

Check protocol metadata, server discovery, HTTP routing headers and session-free migration for agent tool requests.

Validate an MCP 2026 trace →
Long-running tools

MCP Task Trace

Validate durable asynchronous tool lifecycles, input requests, terminal results, polling metadata and migration from experimental Tasks.

Validate MCP Tasks →
Interactive tools

MCP App Ready

Validate the resources, metadata and sandbox policy that let MCP tools render interactive interfaces inside compatible agent hosts.

Validate an MCP App →
Agent UI events

AG-UI Event Lab

Parse raw AG-UI SSE and validate event fields, run IDs, message and tool-call lifecycles, patches, and deprecated event names.

Validate an AG-UI stream →
Agent UI

A2UI Stream Lab

Validate A2UI v0.9 JSON and JSONL messages, surface lifecycle, component references, roots, cycles, and data-model paths.

Validate an A2UI stream →
Instructions

AgentFile Check

Lint AGENTS.md for missing commands, vague requirements, scope ambiguity, contradictions, risky automation, and leaked secrets.

Check AGENTS.md →
Skills

Agent Skill Guard

Inspect SKILL.md frontmatter, instruction overrides, permission requests, hidden downloads, supply-chain behavior, and unsafe examples.

Scan a skill →
Runtime

MCP Config Lab

Translate tool-server configurations among popular coding-agent clients without manually rewriting JSON and TOML.

Convert configuration →
Security

MCP Config Guard

Review the commands, environment variables, packages, and filesystem reach granted to an MCP server.

Review runtime risk →
OAuth

MCP OAuth Lab

Debug metadata discovery and authorization compatibility for remote tool servers used by agents.

Validate OAuth →
Distribution

RegistryReady

Prepare the server.json document that lets developers discover and install an MCP server.

Prepare Registry metadata →
Discovery

ARD Catalog Lab

Generate and validate ai-catalog.json manifests so registries can discover MCP servers, A2A agents, Skills, APIs, and workflows.

Publish an ARD catalog →

What these agent tools check

Agent behavior depends on more than the model. Repository instructions determine how code is changed, skills add reusable workflows, and MCP servers expand access to commands and data. Reviewing those layers together makes failures easier to diagnose and reduces unnecessary permissions.

Designed for explainable review

Every scanner highlights the exact field or instruction behind a warning. That makes the output useful for maintainers, reviewers, and teams establishing an agent governance baseline.