The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes subprocess.run and subprocess.Popen in scripts/dispatch.py to interact with the Feishu (Lark) CLI. These calls are implemented using argument arrays and shell=False, which is a secure practice that prevents shell injection vulnerabilities by ensuring arguments are not interpreted by a command shell.
[DYNAMIC_EXECUTION]: The dispatch function in scripts/dispatch.py executes a command defined by the user in a local config.json file. This mechanism is designed to allow users to connect the task queue to their preferred Agent worker script. The execution is local, documented, and relies on manual configuration by the user, representing normal functional behavior for this tool.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data originating from Feishu document edits and event streams, which constitutes an indirect prompt injection surface. The skill mitigates this risk by including explicit instructions in SKILL.md that direct the Agent to treat all document content as untrusted data and to ignore any embedded instructions that attempt to override owner roles, leak materials, or execute unauthorized commands.

dbs-human-dispatch