The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to install the makeables package globally via npm install --global makeables@latest. Installing software from external registries that do not match the vendor's provided resource naming patterns introduces a dependency on the integrity of the external package.
[COMMAND_EXECUTION]: The skill frequently executes shell commands, including npm install, makeables skills get, and makeables studio start. These commands interact with the local environment, start local servers, and perform network operations such as publishing designs via makeables submit.
[DYNAMIC_EXECUTION]: The skill serves as a 'discovery stub,' where the actual design logic and versioned workflows are not present in the markdown but are fetched at runtime using makeables skills get [guide]. This dynamic loading of instructions from an external CLI tool limits the visibility of the skill's full behavior during static analysis.
[INDIRECT_PROMPT_INJECTION]: The skill's architecture is susceptible to indirect injection because it retrieves and follows instructions from an external source (the CLI output) and user-provided inputs.
Ingestion points: Instructions fetched via makeables skills get commands and user-provided design prompts or images.
Boundary markers: None; there are no delimiters used when the agent processes the output of the CLI commands to distinguish between metadata and instructions.
Capability inventory: Shell execution (npm, makeables CLI), local file system access for saving deliverables and backups, and network access for design submission.
Sanitization: The instructions do not specify any validation or sanitization of the content retrieved from the CLI before the agent interprets it.

makeables