The Agent Skills Directory

[PROMPT_INJECTION]: The skill provides blueprints for creating AI tools and prompts that handle untrusted input. Direct interpolation of raw strings into prompt templates without sanitization creates a vulnerability surface for indirect prompt injection.
Ingestion points: Data enters through request.params.arguments in the code-review and summarize-text prompt handlers.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present in the prompt templates.
Capability inventory: The skill includes code for servers with file system access (fs.readFile, fs.writeFile), network access (axios.get), and database access (pool.query).
Sanitization: While the database example correctly uses parameterized queries, the prompt templates perform direct string interpolation, and the file system example lacks path validation.
[COMMAND_EXECUTION]: The 'File System Server' code template implements file read, write, and directory listing capabilities. The implementation uses path.join to resolve user-supplied file paths without verifying if the resulting path remains within the intended root directory. This pattern is vulnerable to path traversal attacks (e.g., using ../), which could allow an agent to access or modify sensitive files on the host system.
[EXTERNAL_DOWNLOADS]: The skill guides users to install several official and well-known packages from the npm registry, including the official @modelcontextprotocol/sdk, typescript, axios, and pg for database connectivity.

MCP Builder