The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions direct users to install components using npx assistant-ui@latest, which fetches and executes code from the official registry. This is a standard and neutral installation procedure for vendor-provided developer tools.
[INDIRECT_PROMPT_INJECTION]: The skill manages thread lists containing metadata and titles, creating a potential surface for indirect prompt injection where malicious instructions could be embedded in conversation names.
Ingestion points: Thread data is ingested through the RemoteThreadListAdapter (list, fetch methods) and ExternalStoreThreadListAdapter as described in references/remote-adapter.md and references/management.md.
Boundary markers: The documentation and provided UI primitives do not demonstrate the use of specific boundary markers or "ignore" instructions to separate thread titles from agent logic.
Capability inventory: The skill provides network communication capabilities via standard fetch calls to backend endpoints and offers thread lifecycle management (Archive, Delete, Rename) through the aui.threads API.
Sanitization: The implementation relies on standard React rendering for text, which provides protection against script execution (XSS) but does not validate the semantic content of titles against prompt injection.

thread-list