The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes external data in the form of code diffs which could contain adversarial instructions disguised as comments or code slop.
Ingestion points: The skill instructs the agent to "Check the diff against main" as defined in SKILL.md.
Boundary markers: The instructions lack explicit delimiters or "ignore embedded instructions" warnings for the code being analyzed.
Capability inventory: The skill body in SKILL.md implies the agent has permissions to read repository diffs and perform file modification operations to clean up code.
Sanitization: There is no evidence of sanitization, validation, or filtering of the code diff content to prevent the execution of embedded instructions.

cursor-deslop