The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The scripts/iskron-bridge.mjs script performs network requests to https://mcp.iskron.ru to facilitate MCP communication and OAuth 2.1 authentication. This communication is the primary function of the skill and targets the vendor's own infrastructure.
[COMMAND_EXECUTION]: The bridge script uses node:child_process to spawn the system's default browser (using open, start, or xdg-open) to complete the OAuth authentication flow. The skill also provides instructions for the agent to execute setup commands like mkdir and cp to install the bridge script locally.
[PERSISTENCE]: The skill instructions direct the agent to copy the bridge script to ~/.iskron-bridge/ and register it as a global MCP server. This creates a persistent bridge configuration across sessions, which is necessary for the transport layer to function. The script also implements a token keepalive mechanism using setInterval to prevent session expiration.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a proxy for remote MCP tools, introducing a surface where instructions from the remote server could influence the agent.
Ingestion points: scripts/iskron-bridge.mjs ingests JSON-RPC messages from stdin and processes HTTP/SSE responses from mcp.iskron.ru.
Boundary markers: The bridge relies on the structured JSON-RPC protocol to delimit commands and responses.
Capability inventory: The bridge script has network access (node:http, fetch), file system access for token management (node:fs), and process spawning for browser interaction.
Sanitization: The bridge uses JSON.parse to validate incoming messages and follows specific routing logic, which mitigates simple text injection.

establish-mcp