The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads the iskron-bridge executable script and associated plugins/documentation from the vendor's GitHub repository (github.com/iskron-ai/skills).
[REMOTE_CODE_EXECUTION]: The agent is instructed to execute the downloaded JavaScript bridge (iskron-bridge.mjs) using the Node.js runtime. This is the primary function of the skill to enable communication with the remote graph service.
[CREDENTIALS_UNSAFE]: The skill manages authentication by instructing the agent to store personal access tokens in a local file (~/.iskron-bridge/token). It mitigates access risk by using the umask 077 command to ensure the file is created with restricted user-only permissions.
[COMMAND_EXECUTION]: The instructions utilize several shell commands to configure the environment, including directory creation (mkdir), file management (cp, printf), and tool registration (claude mcp add, codex plugin add).
[PERSISTENCE]: The skill establishes persistence for the bridge by adding it to the configuration files of the host application (e.g., claude_desktop_config.json), ensuring the tool remains available in future sessions.
[INDIRECT_PROMPT_INJECTION]: The agent is directed to follow instructions and configuration fixes generated by the iskron-bridge doctor diagnostic command.
Ingestion points: Output from the iskron-bridge doctor and iskron_stand shell commands.
Boundary markers: None provided; the agent is instructed to treat the tool output as authoritative configuration data.
Capability inventory: The agent can write to the file system, modify application configurations, and execute subprocesses via node.
Sanitization: No specific sanitization or validation of the tool's output is mentioned.

establish-mcp