The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill fetches external JSON metadata from the Alibaba Cloud metadata service to generate inventory files and guide agent behavior. This represents an attack surface where maliciously crafted metadata could influence the agent.
Ingestion points: The script scripts/list_openapi_meta_apis.py retrieves data from https://api.aliyun.com/meta/v1/products/hbr/versions/2017-09-08/api-docs.json.
Boundary markers: Absent; the skill relies on the structure of the JSON response from the remote service without explicit delimiters to isolate ingested content from agent instructions.
Capability inventory: The skill can write to the output/ directory, execute Python scripts, and access cloud provider credentials.
Sanitization: The script parses the JSON but does not sanitize the metadata strings before writing them to files or using them in the agent's context.
[EXTERNAL_DOWNLOADS]: The skill retrieves product and API metadata from api.aliyun.com. This is a standard operation for cloud management tools to discover available service capabilities.
[CREDENTIALS_UNSAFE]: The skill instructions explicitly direct the agent to utilize credentials stored in standard Alibaba Cloud configuration locations, including environment variables like ALIBABACLOUD_ACCESS_KEY_ID and the shared credentials file located at ~/.alibabacloud/credentials.

aliyun-hbr-backup