The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The moon tool supports inheriting task configurations from remote URLs using the extends field, as documented in references/task-config.md. This enables the sharing of centralized build logic across repositories via services like GitHub.
[DYNAMIC_EXECUTION]: The tool supports loading and executing WASM-based plugins. As described in references/workspace-config.md, these plugins can be fetched from remote URLs or local file paths to extend the tool's functionality.
[EXTERNAL_DOWNLOADS]: CI configuration examples in examples/ci-workflow.yml include instructions to download and execute toolchain installation scripts from the official moonrepo.dev domain using shell pipes.
[COMMAND_EXECUTION]: Task definitions utilize command and script fields to execute shell operations. The script field specifically supports complex shell features like pipes, redirects, and chaining, which expands the tool's execution capabilities.
[INDIRECT_PROMPT_INJECTION]: The skill documents an architecture where the agent reads and executes instructions from project configuration files, creating a surface for indirect prompt injection if the agent interacts with untrusted repositories.
Ingestion points: Configuration files such as moon.yml and .moon/tasks/*.yml, as well as remote configurations loaded via the extends keyword.
Boundary markers: The documentation does not specify the use of delimiters or instructions to ignore embedded commands within the configuration data.
Capability inventory: The tool possesses broad shell execution capabilities through its task orchestration system, as documented in the configuration references.
Sanitization: No sanitization or validation of the provided command or script strings is described prior to their execution in the system shell.

moon