The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill references official installation scripts using piped shell commands (curl | sh and iwr | iex) from https://open.longbridge.com.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external content from news and community topics, creating an indirect prompt injection surface.
Ingestion points: longbridge news, longbridge topic, ctx.news(), ctx.topics(), and longbridge.com/en/news.md.
Boundary markers: No specific delimiters or instructional guardrails are described for the agent when processing this data.
Capability inventory: Includes network access and CLI command execution.
Sanitization: No sanitization logic for fetched content is mentioned in the references.

stock