The Agent Skills Directory

[COMMAND_EXECUTION]: The skill orchestrates the yike CLI tool to perform media operations, account management, and configuration tasks. It relies on executing system commands with various parameters, including user-supplied prompts and file paths.
[EXTERNAL_DOWNLOADS]: The skill includes instructions for updating the CLI tool (yike update) and installing additional bundled skill packages (yike self skill install). These operations involve fetching and installing code from the vendor's infrastructure.
[INDIRECT_PROMPT_INJECTION]: The skill processes external, untrusted data which creates a potential surface for indirect prompt injection.
Ingestion points: The agent ingests local file paths, Agent chat attachments, and external URLs provided by the user (SKILL.md, references/upload.md).
Boundary markers: The skill uses structured command-line flags (e.g., --prompt, --reference-image) to separate input types, although it does not define specific prompt-level delimiters.
Capability inventory: The tool can read local files, resolve paths, execute CLI commands, communicate with the Yike cloud API, and install additional agent skills (SKILL.md, references/account-and-setup.md).
Sanitization: The instructions mandate verifying file existence, readability, and extension support (e.g., png, jpg, mp4) before proceeding with upload or generation (references/upload.md).
[CREDENTIALS_UNSAFE]: The skill manages authentication using the YIKE_API_TOKEN environment variable. It follows security best practices by explicitly prohibiting the agent from reading browser cookies or profiles and instructing it never to output the token in plaintext (references/account-and-setup.md).

alibabacloud-yike-cli