The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it retrieves and processes untrusted data from the internet via search snippets and full-page extraction.
Ingestion points: The search and extract commands in all CLI scripts (anysearch_cli.py, anysearch_cli.js, anysearch_cli.sh, anysearch_cli.ps1) ingest content from external URLs.
Boundary markers: The skill does not implement explicit delimiters or instruction-ignore headers when passing extracted web content back to the agent.
Capability inventory: The skill can write to local files (.env, runtime.conf) and perform network requests to the provider's API.
Sanitization: No sanitization or filtering of the retrieved HTML/Markdown content is performed before presentation to the AI agent.
[COMMAND_EXECUTION]: The PowerShell implementation (anysearch_cli.ps1) uses -ExecutionPolicy Bypass to run, which is a common but noteworthy technique for bypassing local script execution restrictions.
[CREDENTIALS_UNSAFE]: The skill manages API keys and includes an 'auto-registration' feature where the provider's API can return a new api_key in a response, which the agent is then instructed to persist into the local .env file.
[COMMAND_EXECUTION]: The Python script (anysearch_cli.py) accesses the Windows Registry (HKEY_CURRENT_USER\Software\Microsoft\Windows\CurrentVersion\Internet Settings) to programmatically detect system proxy settings (e.g., for Clash or V2Ray environments).

anysearch