The Agent Skills Directory

[SAFE]: The skill performs legitimate incident response tasks for Vast.ai GPU instances using official tools and documented procedures.
[COMMAND_EXECUTION]: The skill uses vastai, ssh, and aws CLI tools to triage outages, inspect logs, and provision replacement instances. These tools are required for cloud infrastructure management.
[DYNAMIC_EXECUTION]: Uses python3 -c for JSON parsing of CLI output and subprocess.run to automate instance destruction during billing emergencies. This is a common and safe pattern for system automation.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources via vastai logs and vastai show instances.
Ingestion points: vastai logs, vastai show instances --raw (SKILL.md).
Boundary markers: None explicitly defined in the provided script pipes.
Capability inventory: Bash(vastai:*), Bash(ssh:*), Bash(curl:*) allowed in frontmatter.
Sanitization: Uses standard json parsing to handle structured data. The data is used for status assessment and does not present an immediate security risk.

vastai-incident-runbook