longbridge-technical

Warn

Audited by Gen Agent Trust Hub on Oct 12, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill requires several third-party Python packages to be installed via pip, including 'czsc' and 'smartmoneyconcepts'. These are niche, domain-specific libraries that are not from trusted or well-known organizations and should be reviewed for safety.
  • [DYNAMIC_EXECUTION]: The Elliott Wave analysis framework instructs the agent to execute an external script 'scripts/signal_engine.py'. Since the source code for this script was not provided in the skill files, its logic and safety cannot be verified.
  • [DATA_EXPOSURE]: The skill uses the 'longbridge assets' and 'longbridge positions' commands to access sensitive financial information, including the user's total net assets and current stock holdings. This data is used locally to calculate position sizes for the Turtle Trading system, but users should be aware of this access. No evidence of unauthorized network exfiltration was detected.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from market news ('longbridge news') and regulatory filings ('longbridge filing'). This creates a surface for indirect prompt injection if malicious instructions are embedded in the news or filings that the agent then processes for context.
  • Ingestion points: News and filing data fetched at runtime in 'references/elliott-wave.md' and 'references/turtle-signal.md'.
  • Capability inventory: Shell command execution via Longbridge CLI and Python code execution.
  • Boundary markers: None explicitly defined to separate untrusted data from instructions.
  • Sanitization: No specific sanitization or filtering logic is described for the news content.
  • [COMMAND_EXECUTION]: The skill relies extensively on executing shell commands via the 'longbridge' CLI tool to fetch market data and account information. While functional, this depends on the local installation and integrity of the Longbridge terminal environment.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 12, 2026, 12:25 PM