The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The script scripts/bootstrap-cloudflare-start.sh executes npm create cloudflare@latest and npm create @tanstack/start@latest. These commands download and run remote scaffolding scripts from well-known technology organizations (Cloudflare and TanStack) during the project initialization phase.\n- [EXTERNAL_DOWNLOADS]: The skill documentation and scripts facilitate the installation of numerous Node.js packages from the @tanstack and @cloudflare organizations. These are official packages from established tech companies.\n- [PROMPT_INJECTION]: An indirect prompt injection surface is documented in references/devtools-and-llm-support.md. The skill describes sharing route metadata with LLM agents, which could be exploited if an attacker can influence those metadata fields.\n
Ingestion points: Route metadata (titles/descriptions) in app/routes/** files are serialized for agent consumption.\n
Boundary markers: No explicit boundary markers are defined in the examples to isolate data from instructions.\n
Capability inventory: The agent is granted Bash, Read, Write, and Edit tools, which could be abused if the agent follows malicious instructions hidden in metadata.\n
Sanitization: The skill identifies this risk and recommends developers use authentication checks to prevent unauthorized data from reaching the metadata fields.

tanstack-start