The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to direct prompt injection. The research prompt provided by the user is interpolated directly into system and user messages for the AI agents in src/planning_agent.py and src/agents.py. This allows a user to potentially bypass or override the agent's research logic.
[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection.
Ingestion points: Untrusted user input enters the agent context through the prompt field in the /generate_report endpoint defined in main.py.
Boundary markers: There are no delimiters or instructions used to isolate the user-provided prompt from the agent's instructions.
Capability inventory: The agent has the capability to perform network requests to external APIs (Tavily, arXiv) and perform read/write operations on a PostgreSQL database.
Sanitization: No sanitization or validation is applied to the user input before it is passed to the LLM.
[CREDENTIALS_UNSAFE]: The skill's documentation and code (specifically main.py) use hardcoded default credentials (postgresql://app:local@...) and default environment variable values for the database. While these are for a local development environment, hardcoded credentials represent a security best-practice violation.
[EXTERNAL_DOWNLOADS]: The installation instructions include cloning a repository from DeepLearning.AI's official GitHub organization. This is a reference to a well-known source and is documented as part of the setup process.

agentic-research-agent-deeplearning