The Agent Skills Directory

[DATA_EXFILTRATION]: The skill directs network requests to api.minimaxi.com and api.minimaxi.chat. These domains are typosquatted variations of the official MiniMax AI domains (e.g., minimax.chat), which is a characteristic pattern for exfiltrating sensitive credentials to attacker-controlled infrastructure.- [CREDENTIALS_UNSAFE]: The skill instructs users to store a MINIMAX_API_KEY in their environment and then transmits this secret to the aforementioned suspicious domains.- [PROMPT_INJECTION]: The skill retrieves and processes untrusted data from the web, exposing the agent to indirect prompt injection attacks.\n
Ingestion points: Search results (titles, snippets) fetched in scripts/search.ts.\n
Boundary markers: No delimiters or safety instructions are used to isolate the retrieved content.\n
Capability inventory: The script performs network requests and outputs data to the system console.\n
Sanitization: The retrieved content is not sanitized or validated before processing.- [REMOTE_CODE_EXECUTION]: The script scripts/search.ts imports core functionality from ../../../scripts/vendor/minimax-core. This relative import points to code outside the skill's own directory, making the full execution logic unverified within the scope of the individual skill package.

minimax-search