The Agent Skills Directory

[PROMPT_INJECTION]: The skill generates code for a RemoteAnnouncementProvider that fetches JSON data from an external URL to populate banner text. This creates a surface for indirect prompt injection if the remote data source is compromised.
Ingestion points: The RemoteAnnouncementProvider.fetchAnnouncements() method in the generated code ingests JSON data from a remote endpoint.
Boundary markers: The implementation does not include specific boundary markers or delimiters for the fetched content to prevent it from being interpreted as instructions by the agent.
Capability inventory: The skill requires tools like Bash, Write, and Read to generate the project files. However, the generated code itself is limited to UI presentation, network fetching, and local storage.
Sanitization: The generated code uses standard JSONDecoder for parsing but does not include explicit sanitization or filtering of the fetched message strings.

announcement-banner