The Agent Skills Directory

[COMMAND_EXECUTION] (HIGH): The file auto_scan.js uses execSync to call an external script with arguments derived from previous command outputs. Specifically, the variable group.chat_id is interpolated directly into a shell command string: node "${GROUP_INTEL_SCRIPT}" members "${group.chat_id}". If the data returned by the group list command contains shell metacharacters in the chat_id field, it could lead to arbitrary command execution.
[PROMPT_INJECTION] (HIGH): Indirect Prompt Injection Surface (Category 8). The skill ingests untrusted data from the Feishu API and other scripts to populate a persistent identity registry.
Ingestion points: sync.js (Feishu API) and auto_scan.js (external script output).
Boundary markers: None. Data is parsed and written directly to user_registry.json.
Capability inventory: execSync in auto_scan.js, file-write operations to user_registry.json in index.js, sync.js, and auto_scan.js.
Sanitization: Absent. There is no escaping or validation of user-controlled strings (names, IDs, aliases) before they are stored or used in command line arguments.
[EXTERNAL_DOWNLOADS] (LOW): The skill depends on fs-extra, a common and trusted Node.js package. No suspicious remote code downloads were detected.

identity-manager