longbridge-content

Pass

Audited by Gen Agent Trust Hub on Oct 12, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, including news articles and community discussions, which presents a surface for indirect prompt injection.
  • Ingestion points: External data is ingested via CLI commands such as longbridge news and longbridge topic as documented in references/news.md and references/topic.md.
  • Boundary markers: The skill employs structured output templates and classification buckets to organize findings, although it does not use explicit XML-style delimiters for raw data.
  • Capability inventory: Capabilities are limited to financial data retrieval, narrative analysis of filings, and the ability to post replies to community topics. The skill does not have access to sensitive local files, environment variables, or arbitrary network execution.
  • Sanitization: The skill contains explicit instructions to detect and downgrade 'hype vocabulary' and speculative trading terms (e.g., 'all in', '庄家') found in community content, preventing the agent from repeating potentially manipulative or low-quality external instructions.
  • [COMMAND_EXECUTION]: The skill utilizes the longbridge CLI tool to perform its primary functions. These commands are vendor-owned and aligned with the skill's stated purpose of providing content from the Longbridge platform.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 12, 2026, 12:25 PM