The Agent Skills Directory

[EXTERNAL_DOWNLOADS] (LOW): The scripts/init-mcp-server.sh script executes npm install to download and install project dependencies from the npm registry. This is expected behavior for a project initialization script.
[COMMAND_EXECUTION] (LOW): The skill includes shell scripts (init-mcp-server.sh, test-mcp-connection.sh) that use command-line tools like sed and curl to scaffold projects and verify endpoint connectivity.
[INDIRECT_PROMPT_INJECTION] (LOW): The skill exposes surfaces for indirect prompt injection through tools that ingest and process external data.
Ingestion points: Tool arguments in templates/full-server.ts and templates/tool-server.ts (e.g., query, table, text) and responses from external APIs (e.g., weather data).
Boundary markers: Absent in the provided templates; developers are encouraged to implement their own.
Capability inventory: Includes network access via fetch, database operations via D1, and key-value storage via KV.
Sanitization: Uses Zod for basic schema validation and parameterized queries for most database interactions. However, templates/full-server.ts contains a pattern that interpolates a table name directly into a SQL query string (SELECT * FROM ${table}), which is a potential security risk if the input is not strictly validated.

typescript-mcp