The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external plugin code provided via the $ARGUMENTS parameter. This creates a surface for indirect prompt injection where malicious instructions embedded in the analyzed code could influence the agent's behavior during the modification process.\n
Ingestion points: Source files (.ts, .tsx) read from the directory or file path specified in $ARGUMENTS.\n
Boundary markers: Absent. The skill does not instruct the agent to distinguish between its own instructions and content found within the target source files.\n
Capability inventory: The agent is granted Read, Write, Edit, and Bash tools, providing a significant impact surface if an injection occurs.\n
Sanitization: The skill does not perform validation or sanitization of the input code before analyzing or editing it.

background-state