The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill uses pip to install and upgrade the vastai package from PyPI. This involves downloading and executing code from an external repository to update the environment tools.
[COMMAND_EXECUTION]: The skill provides instructions for executing various system-level commands, including docker pull and docker run for container management, as well as several vastai CLI operations for searching offers and managing instances.
[DYNAMIC_CONTEXT_INJECTION]: The skill uses the !command syntax in the SKILL.md file to execute vastai --version and pip show vastai at load time. These commands are used to automatically detect the current state of the installation and inform the agent's context.
[DYNAMIC_EXECUTION]: The skill generates and executes code at runtime, such as the get_cli_commands Python function in Step 2 used to parse tool help text, and the verification shell script in Step 5.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external CLI and API sources, creating a surface for potential indirect prompt injection attacks.
Ingestion points: Data enters the context from the output of vastai --help (parsed in Step 2) and JSON data from vastai search (parsed in Step 5).
Boundary markers: The instructions do not define specific delimiters or instructions for the agent to ignore embedded commands within the processed tool outputs.
Capability inventory: The skill has capabilities to execute shell commands (Bash) and write files, providing a path for potential exploitation if external data is malicious.
Sanitization: The skill utilizes basic string manipulation and standard JSON parsing but lacks specific security sanitization or validation to prevent instructions in data from influencing agent behavior.

vastai-upgrade-migration