The Agent Skills Directory

[PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill is designed to store and retrieve arbitrary text data (knowledge items) which are provided by other components in the agent ecosystem. This creates a surface where malicious instructions could be stored and later retrieved to influence the agent's behavior.
Ingestion points: The KnowledgeBase.add method ingests content, title, and metadata from external skill invocations into SKILL.md logic.
Boundary markers: Absent. The skill stores and returns raw text strings without adding delimiters or 'ignore' instructions.
Capability inventory: This skill itself possesses no dangerous capabilities; it lacks network access, shell command execution, and sensitive file access. It strictly performs database operations and mathematical calculations.
Sanitization: While the skill correctly uses parameterized queries to prevent SQL injection, it does not perform semantic sanitization or escaping of the stored content.

knowledge-base