The Agent Skills Directory

[COMMAND_EXECUTION]: The skill orchestrates the execution of various arkcli subcommands to perform tasks such as identity verification, profile switching, and resource resolution. These commands are fundamental to the skill's role as a CLI interface.
[EXTERNAL_DOWNLOADS]: It provides instructions for installing the official @volcengine/ark-cli package via npm and describes a process for downloading the arkexp utility from the vendor's official content delivery network. These actions are performed using standard vendor-managed channels.
[CREDENTIALS_UNSAFE]: The skill handles sensitive authentication artifacts including API keys, session tokens (STS), and Access Key/Secret Key (AK/SK) pairs. It explicitly includes guidelines to prevent the exposure of these secrets in plain text and manages them via its internal profile system.
[INDIRECT_PROMPT_INJECTION]: The skill exposes a potential attack surface by processing user-supplied data such as model names and endpoint IDs into CLI execution strings.
Ingestion points: User parameters provided for commands like arkcli profile use and arkcli resources resolve in SKILL.md and references/execution-context.md.
Boundary markers: Implements structured decision-making logic (0/1/N selection) to disambiguate user intent before executing operations.
Capability inventory: Extensive use of shell execution via the arkcli binary for managing cloud resources and authentication states.
Sanitization: Employs fail-fast mechanisms for flag validation and requires explicit human-in-the-loop confirmation for high-risk operations like resource deletion.

arkcli-shared