The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the user to install dependencies using a piped command that downloads and executes a shell script directly from the author's GitHub repository.
Evidence: curl -fsSL https://raw.githubusercontent.com/yschimke/compose-ai-tools/main/scripts/install.sh | bash in SKILL.md.
[EXTERNAL_DOWNLOADS]: Fetches configuration schemas and references scripts hosted on external domains controlled by the author.
Evidence: Links to https://raw.githubusercontent.com/yschimke/compose-ai-tools/main/scripts/design-artifacts/catalog.spec.schema.json and references to the yschimke/compose-ai-tools and yschimke/design-parity repositories.
[COMMAND_EXECUTION]: Invokes multiple Node.js scripts and a custom command-line interface to render code and manage design specifications.
Evidence: Calls to node scripts/design-artifacts/init-catalog-spec.mjs, node scripts/design-artifacts/validate-catalog-spec.mjs, and the compose-preview CLI.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because it ingests and processes untrusted data from local project files and source code.
Ingestion points: Reads from catalog.spec.json and scans Kotlin source files for @Preview annotations.
Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are present in the processing logic.
Capability inventory: The skill can execute shell commands (node, compose-preview) and perform file system writes through its export libraries.
Sanitization: No evidence of data sanitization or validation of the content within the processed Kotlin files or the JSON specification is provided in the instructions.

compose-design-catalog