The Agent Skills Directory

Indirect Prompt Injection (LOW): The skill ingests user queries and processes external CSV data files. This creates a vulnerability surface for indirect prompt injection where malicious instructions in the database or query could attempt to influence the agent. However, the risk is negligible as the skill lacks side-effect capabilities such as file writing, network access, or command execution.
Ingestion points: User-supplied query argument in search.py and CSV data files located in the data/ directory.
Boundary markers: None present.
Capability inventory: No subprocess calls, file-write operations, or network requests are performed by core.py or search.py.
Sanitization: Input is tokenized by removing punctuation and lowercasing for search indexing, but no instruction-specific sanitization is implemented.
Command Execution (SAFE): Analysis of core.py and search.py confirms the absence of shell-executing functions like os.system or subprocess.run.
External Downloads (SAFE): The skill does not perform any network operations or remote code fetching. Documentation in SKILL.md regarding Python installation is informational for the user and not an automated action.

ui-ux-design