The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill transmits sensitive authentication data over an insecure channel. In scripts/search.py, the VIKING_KBSVR_API_KEY (Bearer token) is included in headers sent to an http:// URL instead of https://, exposing credentials to potential interception.
[COMMAND_EXECUTION]: The skill executes a local Python script scripts/search.py using arguments provided by the agent to perform knowledge base queries.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of volcengine, aiohttp, and requests packages. These are standard libraries for interacting with the targeted service.
[DATA_EXFILTRATION]: The skill communicates with api-knowledgebase.mlp.cn-beijing.volces.com. This is an official Volcengine (Bytedance) domain, consistent with the author's identity. It uses environment variables for authentication.
[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection by processing external knowledge base content.
Ingestion points: scripts/search.py retrieves search results from the Volcengine API and prints the content directly to the agent's context.
Boundary markers: No delimiters or instructions to ignore embedded commands are used when outputting retrieved content.
Capability inventory: The skill possesses network communication capabilities via the requests library.
Sanitization: The content retrieved from the knowledge base is not validated, escaped, or filtered before being presented to the agent.

viking-knowledgebase