The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes natural language requests and voice transcripts to initiate setup, creating a surface area for indirect prompt injection.
Ingestion points: Processes user input and voice transcripts as described in SKILL.md.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present in the processing logic.
Capability inventory: The skill possesses significant capabilities, including writing to AI client configuration files (e.g., .claude.json, .cursor/mcp.json), spawning child processes using node:child_process (found in scripts/lib/process.mjs), and performing network requests via node:net and node:http (found in scripts/lib/smoke.mjs and scripts/lib/token-web.mjs).
Sanitization: Sensitive tokens are passed to child processes via standard input rather than command-line arguments, and a redactSecret function in scripts/lib/shared.mjs is used to filter secrets from logs and output.
[COMMAND_EXECUTION]: The skill executes shell commands to configure and test the MCP server environment.
Executes npx -y mcp-yandex-kit@latest as the primary server command.
Provides a native-configure utility in scripts/lib/process.mjs that can execute arbitrary CLI commands for client compatibility, although instructions limit this to official vendor tools.
Performs chmod 0600 on configuration files to secure sensitive tokens, which is a security best practice.
[EXTERNAL_DOWNLOADS]: The skill makes network connections to external services to fulfill its purpose.
Connects to registry.npmjs.org via npx to fetch the mcp-yandex-kit package.
Connects to api.kit.yandex.net to perform store validation during the smoke test (scripts/lib/smoke.mjs).
These connections target well-known or vendor-owned services related to the skill's primary function.

a1-yandex-kit-setup