The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The workflow executes the @just-every/crawl package using npx -y. This downloads and runs external code from a non-trusted scope on the NPM registry at runtime.
[EXTERNAL_DOWNLOADS]: Fetches model documentation from https://help.aliyun.com/zh/model-studio/models during the crawling step.
[COMMAND_EXECUTION]: Automates a series of shell operations including the installation and execution of Node.js tools, Python scripts for data processing, and filesystem operations like directory creation.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes untrusted data crawled from a public website.
Ingestion points: The crawler saves external web content into alicloud-model-studio-models.md, which is then consumed by the processing scripts.
Boundary markers: No explicit delimiters or instructions are used to isolate the crawled content from the agent's logic.
Capability inventory: The skill possesses the ability to write to the local filesystem and execute shell commands.
Sanitization: The Python scripts use basic regular expressions to extract data, which does not provide robust protection against malicious payloads embedded in the crawled markdown.

aliyun-modelstudio-crawl-and-skill