The Agent Skills Directory

[COMMAND_EXECUTION]: Instructs the agent to perform local environment discovery using ls -la on backend and frontend feature directories to analyze existing implementation details.
[COMMAND_EXECUTION]: Directs the agent to manage local development services and databases using npm run dev, docker-compose, and npm run prisma:migrate.
[COMMAND_EXECUTION]: Provides instructions for verifying API endpoints using curl against http://localhost:3000, including bypassing authentication for local development using the RAPID_LOCAL_DEV=true environment variable.
[PROMPT_INJECTION]: Detected an Indirect Prompt Injection surface where malicious instructions in processed files could influence agent behavior.
Ingestion points: Analyzes directory structures (ls -la) and file content within backend/src/api/features/ and frontend/src/features/ to create implementation plans.
Boundary markers: Uses specific plan templates but lacks explicit instructions or delimiters to ignore embedded natural language commands found in the source code files.
Capability inventory: The agent can execute shell commands (npm, docker-compose, curl), perform file system operations, and generate code/plans based on analyzed content.
Sanitization: No validation or sanitization of the content read from existing project files is performed before inclusion in the reasoning process.
Mitigation: Includes a mandatory human checkpoint ("STOP and wait for 'Go' or 'Proceed' from user") before the agent moves from planning to code implementation.
[EXTERNAL_DOWNLOADS]: References standard package management workflows via npm for installing dependencies and running development scripts.

plan-backend-frontend