The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user code and review instructions, which serves as a surface for indirect prompt injection attacks where malicious code could attempt to override agent instructions.
Ingestion points: SKILL.md defines input parameters for code snippets and review rules which are directly processed by the agent.
Boundary markers: The instructions lack explicit boundary markers or delimiters to differentiate between the agent's instructions and the untrusted user code being analyzed.
Capability inventory: The skill utilizes Language Server Protocol (LSP) tools and Grep for deep code analysis and symbol searching across the file system.
Sanitization: There is no evidence of sanitization or filtering of user-supplied code before it is interpreted by the agent.

ascendc-operator-code-review