The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a bundled Python script scripts/upload_to_skillhub.py to process local files and communicate with the remote API.
[CREDENTIALS_UNSAFE]: The skill relies on sensitive environment variables (ARK_SKILL_API_KEY) for authentication. These are accessed using standard os.getenv methods and used as Bearer tokens in HTTP headers.
[PROMPT_INJECTION]: The SKILL.md contains directives to the agent (e.g., "严禁向用户询问参数值"
Strictly forbidden to ask the user for parameter values) to ensure it operates autonomously using the pre-injected environment variables without exposing or confirming them with the user.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface that allows it to read and upload arbitrary local directories or files specified by the agent.
Ingestion points: Files and directories at local paths specified by the --path argument.
Boundary markers: None identified.
Capability inventory: Reads local filesystem contents, creates temporary zip archives, and performs HTTP POST requests to a remote endpoint.
Sanitization: Filenames are sanitized to remove non-alphanumeric characters, but the content of the files is uploaded without inspection.

byted-skillhub-upload