The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it reads and processes external source files (modules/**/*.provider.ts) to determine implementation details. Malicious instructions or malicious package recommendations embedded in comments within these files could influence the agent to perform unauthorized actions or introduce vulnerabilities.
Ingestion points: Local domain provider interface files and their related types.
Boundary markers: None found; the skill reads file content directly into context.
Capability inventory: Writing implementation files, modifying NestJS modules, and executing shell commands for package installation and testing.
Sanitization: No validation or sanitization of the input file content is performed.
[REMOTE_CODE_EXECUTION]: The skill is authorized to install external libraries using npm install. It suggests the use of standard, well-known libraries such as bcrypt, jsonwebtoken, and nodemailer for technical implementations. While the NPM registry is a well-known service, this capability allows for the introduction of external code into the runtime environment.
[COMMAND_EXECUTION]: The workflow involves running shell commands to manage dependencies within the @poupig/backend workspace and executing project build and test scripts.
[EXTERNAL_DOWNLOADS]: The skill facilitates downloading and integrating third-party dependencies from the NPM registry.

backend-provider-implementation