The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external data for visualization, creating a surface for potential injection attacks.\n
Ingestion points: The instructions guide the agent to analyze and test with real data, which would typically be loaded via the Read or Grep tools.\n
Boundary markers: There are no explicit instructions to treat external data as untrusted or to use delimiters to separate data from the agent's internal logic.\n
Capability inventory: The skill allows access to powerful tools like Bash for command execution and Write/Edit for modifying the file system.\n
Sanitization: The implementation steps do not include requirements for sanitizing or validating the data before it is passed to visualization libraries.\n- [OBFUSCATION]: The provided Python examples include unnecessarily complex code patterns that appear intended to obscure simple operations.\n
Evidence: In Example 1, the script uses .__class__.__call__ to invoke a dictionary constructor. While the resulting operation is benign, the use of indirect attribute access for standard constructors is a common obfuscation technique used to bypass simple keyword-based security scanners.

data-visualization