The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to use npx to download and execute the @derived-modular/cli package from the NPM registry. It also facilitates the installation of related linter plugins such as @derived-modular/eslint-plugin, @derived-modular/oxlint-plugin, and @derived-modular/biome-plugin to support architecture enforcement.
[REMOTE_CODE_EXECUTION]: The use of npx to run the DMA CLI constitutes remote code execution, as the tool is fetched and run in the local environment to perform project analysis and modification.
[COMMAND_EXECUTION]: The skill executes several command-line operations (dma init, dma promote, dma check, dma doctor) that interact with the local filesystem. These commands are used to create new configuration files, add scripts to package.json, move module directories (e.g., from features/ to services/), and programmatically rewrite source code import literals.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because its automated refactoring actions are driven by the analysis of untrusted project files.
Ingestion points: The skill reads the project's source directory (src/), directory structure, and various configuration files such as dma.config.*, .eslintrc*, biome.json, and .oxlintrc.json.
Capability inventory: The CLI tool can move files and modify source code based on its analysis of the project's internal import graph.
Boundary markers: No explicit delimiters or safety instructions are used to distinguish untrusted project data from the skill's operational logic, which could allow maliciously crafted code patterns to influence refactoring decisions.
Sanitization: The skill does not appear to validate or sanitize the analyzed project files before they are used to determine how files should be moved or how code should be rewritten.

dma