The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted application UI content as part of its polish pass.
Ingestion points: The skill is instructed to "Walk every screen and every state" and "Drive the running app" (SKILL.md), ingesting the application's DOM and content.
Boundary markers: No explicit boundary markers or instructions to ignore embedded commands within the UI are provided.
Capability inventory: The skill has the capability to modify application code (CSS/JS) and microcopy, and to perform browser interactions (SKILL.md).
Sanitization: The instructions do not mention sanitizing or escaping the data ingested from the application.
[COMMAND_EXECUTION]: The file evals/run-static-checks.sh is a shell script that executes a local Python script for static validation of the skill.
[EXTERNAL_DOWNLOADS]: The skill metadata (skill.json) references design and usability resources from well-known organizations including Google, Apple, and the Nielsen Norman Group.

moonshot-ux