The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The skill utilizes dynamic context injection syntax (!grep...) in SKILL.md to perform architectural linting and detect anti-patterns such as duplicate entity names or direct cross-module imports. These commands are executed locally on the project source code and do not involve network operations or access to sensitive system directories.
[COMMAND_EXECUTION]: The skill includes scripts/validate-isolation.sh, a shell script designed for static analysis of the codebase. It uses standard utilities like grep and awk to verify module boundaries and naming conventions. No dangerous operations or privilege escalations were detected in the script.
[INDIRECT_PROMPT_INJECTION]: The skill operates by ingesting user-supplied domain information to design and implement backend modules (documented in SKILL.md Phases 1 and 2). It actively mitigates injection risks by enforcing the use of Data Transfer Objects (DTOs) with class-validator and class-transformer for all inputs, and recommending strict TypeScript configurations as described in references/architecture-patterns.md.
[EXTERNAL_DOWNLOADS]: The skill references standard software packages from the official NPM registry, including @thallesp/nestjs-better-auth, kafkajs, ioredis, and AWS SDK components. These are well-known libraries within the NestJS ecosystem.
[CREDENTIALS_UNSAFE]: The skill correctly handles sensitive information by explicitly instructing the user to avoid hardcoded configurations and to manage secrets like DATABASE_URL and JWT_SECRET via environment variables, as seen in references/authentication.md and references/stack-configuration.md.

nestjs-modular-monolith