The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill points to installation resources hosted on the author's official GitHub organization ("volcengine/ve-storage-uni-cli").
[REMOTE_CODE_EXECUTION]: Documentation suggests a standard installation method involving downloading a shell script from the vendor's GitHub releases and piping it to a shell interpreter.
[COMMAND_EXECUTION]: The skill executes the tos-cli command-line interface to perform storage operations such as listing, status checking, and synchronization.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection due to its interaction with external object storage.
Ingestion points: Output from tos-cli ls and tos-cli stat is ingested into the agent context, potentially containing untrusted metadata or file names.
Boundary markers: No explicit delimiters are used in the prompt instructions to isolate external tool output from the agent's core logic.
Capability inventory: The skill possesses the capability to perform data transfers (cp, sync) and generate presigned URLs.
Sanitization: The skill provides a safety reference (references/safety.md) that encourages the use of --dry-run and path quoting to prevent accidental command execution or data exposure.

tos-cli