The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is implemented as a Python CLI tool executed via uv run in the skill's scripts directory to handle storage and CDN operations locally.
[DATA_EXFILTRATION]: The skill performs network operations to upload local file contents to remote S3-compatible storage endpoints and interacts with Tencent Cloud CDN APIs for cache management.
[CREDENTIALS_UNSAFE]: The skill manages S3 and CDN credentials (Access Key ID and Secret Access Key) via configuration files and environment variables. The implementation uses best practices such as SecretValue objects with repr=False to minimize the risk of accidental credential exposure in terminal output or logs.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection: \n
Ingestion points: The upload command accepts a local_path argument provided by the agent. \n
Boundary markers: None; the skill does not implement internal restrictions or markers to distinguish between safe and sensitive local file paths. \n
Capability inventory: The skill has the capability to read any local file (subject to agent permissions) and transmit it to an external network endpoint. \n
Sanitization: The skill prevents path traversal (..) in the remote S3 object keys but does not validate or sanitize the local source path against sensitive directories.

object-storage