The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses npx ai-devkit@latest, which instructs the agent to download and execute the newest version of the ai-devkit package from the public NPM registry during operation.
[REMOTE_CODE_EXECUTION]: By invoking npx, the skill executes code sourced from a remote repository at runtime. This dependency on external code that is not pinned to a specific version or hash introduces risks if the remote package is compromised.
[COMMAND_EXECUTION]: The skill executes shell commands to perform linting and validation. There is a potential for command injection if the <name> placeholder for features is populated with unvalidated user input containing shell metacharacters.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it is designed to read and process existing requirements and design documents from the file system.
Ingestion points: Existing requirements and design docs found in docs/ai or paths defined in .ai-devkit.json.
Boundary markers: None specified in the instructions to distinguish between instructions and processed data.
Capability inventory: Shell command execution via npx, file-system read access for documentation, and file-system write access for updating design docs.
Sanitization: No explicit sanitization or validation of the content read from external documents is mentioned.

dev-design