The Agent Skills Directory

[PROMPT_INJECTION]: The skill implements defensive measures against indirect prompt injection by explicitly instructing the agent in the discovery, creation, and validation workflows to ignore any embedded directives found in fetched pages, user briefs, or wireframes.
[EXTERNAL_DOWNLOADS]: The skill utilizes the WebFetch tool to retrieve the structural elements of user-provided reference URLs. The instructions ensure that only layout information is extracted, and no content is executed.
[SAFE]: The skill demonstrates safe data handling by stripping unnecessary metadata (such as requirement or story IDs) and replacing specific copy strings with abstract labels, ensuring the generated blueprints remain design-blind and free of potentially sensitive external strings.

blueprint