The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external data, including code diffs, repository history, and symbol definitions, which serves as a potential surface for indirect prompt injection.
Ingestion points: The SKILL.md instructions involve reading diffs, repository history, and changed symbols.
Boundary markers: There are no explicit delimiters defined to separate untrusted code content from the agent's instructions.
Capability inventory: The skill is authorized to create and execute local scripts and tests to verify behavior.
Sanitization: No specific filtering or sanitization steps are defined for the analyzed code before it is incorporated into local probes.
[DYNAMIC_EXECUTION]: The skill creates and executes local tests or scripts to 'prove safety facts' by exercising the real code within the local environment. This runtime generation and execution is a core part of the skill's diagnostic function.
[SAFE]: The skill references and adapts logic from the public plugins repository of a well-known AI development platform (Cursor), which is documented neutrally in the third-party notices.

blast-radius