The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of child_process.spawnSync in src/platforms/agentReach.js to run multiple external CLI tools such as agent-reach, opencli, bili, xhs, and mcporter. This represents a high-privilege capability within the agent's environment.
[COMMAND_EXECUTION]: The function runCustomDouyin in src/platforms/agentReach.js executes a command defined by the DOUYIN_COMMAND environment variable. If an attacker can influence environment variables, they could achieve arbitrary command execution.
[REMOTE_CODE_EXECUTION]: The commandExists function in src/platforms/agentReach.js passes a variable directly into a shell string: sh -lc "command -v ${command}". When command is sourced from the user-controlled DOUYIN_COMMAND environment variable, it creates a surface for shell command injection.
[PROMPT_INJECTION]: The skill is vulnerable to Indirect Prompt Injection (Category 8) as it processes untrusted data from external sources and possesses command execution capabilities.
Ingestion points: src/platforms/agentReach.js (Fetches video details, search results, and comments from Bilibili and Guaikei API).
Boundary markers: None identified. Data is interpolated directly into console outputs and JSON results.
Capability inventory: Subprocess execution via spawnSync in src/platforms/agentReach.js.
Sanitization: src/validate/keyword.js performs basic character cleaning for keywords, but does not sanitize the content retrieved from external APIs before outputting or logging.
[CREDENTIALS_UNSAFE]: The skill requires the GUAIKEI_API_TOKEN environment variable for fallback operations. While it follows the standard practice of using environment variables rather than hardcoding, the token is transmitted to www.guaikei.com which is a third-party service.

global-content-search