The Agent Skills Directory

[SAFE]: The skill does not contain any malicious code, obfuscation, or unauthorized data access patterns. It focuses on providing safe implementation guidelines for the TanStack Start framework.
[PROMPT_INJECTION]: No prompt injection or behavior override patterns were detected. The use of the word 'CRITICAL' is contextually appropriate for highlighting security requirements in the framework.
[DATA_EXFILTRATION]: No sensitive file access or unauthorized network exfiltration patterns were found. The examples use standard web framework patterns for database and API interaction.
[REMOTE_CODE_EXECUTION]: The skill references standard, well-known libraries such as @tanstack/react-start and zod. There are no instances of unverified package installations or remote script execution.
[INDIRECT_PROMPT_INJECTION]: The skill describes data ingestion through createServerFn handlers and validators. While this represents a potential attack surface, the skill explicitly instructs users to implement validation and authentication.
Ingestion points: Data is received via the .validator() and .handler() methods of createServerFn in SKILL.md.
Boundary markers: The skill emphasizes the use of Zod or custom function validators to enforce data schemas at the boundary.
Capability inventory: Handlers in the examples demonstrate database queries (db.query), deletions (db.delete), and external fetching (fetchPublicData) in SKILL.md.
Sanitization: Validation is enforced at the entry point, and the documentation warns about the need for authorization and careful cache control.

tanstack-start-client-core-start-core-server-functions