The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to install the mcporter utility globally using npm install -g mcporter and configures the environment to use a private PyPI mirror at https://mirrors.ivolces.com/pypi/simple/.
[REMOTE_CODE_EXECUTION]: The MCP configuration provided in SKILL.md uses uvx to download and execute the Flink MCP server directly from the vendor's GitHub repository at github.com/volcengine/mcp-server.
[COMMAND_EXECUTION]: The skill manages Flink applications (including deployment, starting, stopping, and scaling) by executing shell commands via the mcporter interface.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it reads and processes Flink application logs (list_flink_application_log) and system events. If the Flink job processes untrusted data and logs error details, an attacker could potentially influence the agent's diagnostics or cause it to perform unintended management actions.
Ingestion points: Application logs and events are retrieved in flink-dev.md, flink-diagnosis.md, and flink-resource.md.
Boundary markers: There are no explicit instructions or delimiters used to ensure the agent treats log content strictly as data rather than instructions.
Capability inventory: The skill has access to dangerous management tools such as stop_flink_application, start_flink_application, and update_flink_application_draft.
Sanitization: The instructions do not define any sanitization or filtering logic for external data retrieved from logs.

byted-volcengine-flink