The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external URIs from the operating system, which constitutes a surface for indirect prompt injection from untrusted external sources.
Ingestion points: The agent is instructed to listen to external link events via AppLinks().uriLinkStream and AppLinks().getInitialLink(), as described in references/01-foundations-package-lifecycle.md.
Boundary markers: The instructions mandate a trust boundary where raw URIs are parsed and converted into typed DeepLinkIntent objects before being used by the application, which serves as a delimiter between untrusted input and application logic (defined in SKILL.md Step 8 and references/01-foundations-package-lifecycle.md).
Capability inventory: The skill utilizes navigation capabilities through go_router or Navigator to change the application's view or state based on the external input (detailed in SKILL.md Step 11 and references/03-routing-auth-security.md).
Sanitization: The skill contains extensive instructions for URI sanitization, including the use of the Uri parser, exact host and scheme allowlisting, and validation of path parameters to prevent malicious payloads from influencing the agent's behavior (defined in SKILL.md Step 7 and references/03-routing-auth-security.md).

flutter-app-links-deep-link-engineering