The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documents the run_script configuration field, which allows the AXIS framework to execute arbitrary shell commands during scenario setup and teardown phases.
[EXTERNAL_DOWNLOADS]: The skill describes capabilities for cloning remote git repositories into a scenario workspace via the copy action and fetching agent skills or configuration from GitHub repositories.
[INDIRECT_PROMPT_INJECTION]: The skill defines a vulnerability surface where the framework ingests content from external sources (git repositories, remote skills, and third-party MCP servers).
Ingestion points: External git repositories via copy, remote skills via GitHub URLs, and remote mcp_servers.
Boundary markers: The skill specifies that scenarios run in an "isolated workspace."
Capability inventory: Scenario lifecycle actions include run_script (shell execution) and configurable mcp_servers (filesystem/tool access).
Sanitization: No explicit sanitization of external content is described within the instruction set.
[DYNAMIC_EXECUTION]: The framework supports the use of custom adapters, which are JavaScript or TypeScript modules loaded and executed by the framework at runtime.

configure-axis