The Agent Skills Directory

[SAFE]: The skill provides documentation and commands for testing local MCP servers. All external tools mentioned, such as @modelcontextprotocol/inspector and ajv-cli, are established tools from well-known sources or official organizations.
[COMMAND_EXECUTION]: The skill instructs the agent to run local verification commands (e.g., npx, python) against the user's server code. These are standard development workflows intended for the stated purpose of server verification.
[INDIRECT_PROMPT_INJECTION]: The skill identifies an attack surface where an agent processes data from an MCP server.
Ingestion points: The agent processes tools/list and tools/call JSON-RPC responses from a local server (documented in references/protocol-conformance.md).
Boundary markers: No specific boundary markers or instructions to ignore embedded content are included in the verification guides.
Capability inventory: The skill utilizes npx and python to execute validation tools and suggests shell-based hooks for logging (SKILL.md, references/protocol-conformance.md).
Sanitization: No explicit sanitization or validation of server-provided tool descriptions or content is specified before the agent processes them.

mcp-verification