The Agent Skills Directory

[SAFE]: Extensive analysis of the skill's instructions, metadata, and scripts revealed no malicious patterns or indicators of compromise.\n- [EXTERNAL_DOWNLOADS]: The skill initiates network requests to 'generativelanguage.googleapis.com'. This is a well-known and trusted service provided by Google for the Gemini API, and its use is consistent with the skill's stated purpose of providing grounded search answers.\n- [DATA_EXFILTRATION]: User API keys are handled securely. The code in 'scripts/index.html' passes the 'secret' parameter directly to the official Google API endpoint for authentication. No evidence of credential harvesting or unauthorized transmission to third-party domains was found.\n- [PROMPT_INJECTION]: The instructions in 'SKILL.md' contain defensive measures that require the agent to preserve user intent literally and avoid reinterpreting factual queries. This minimizes the risk of prompt manipulation during the query construction phase.\n- [COMMAND_EXECUTION]: Tool usage is limited to the standard 'run_js' function with a static local file ('index.html'). The script contains no logic for arbitrary command execution or system-level modifications.

gemini-search