solana-sniper-bot
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources and interpolates it directly into a prompt sent to an LLM for rugpull risk assessment. This creates a vulnerability where external data could influence the LLM's logic.\n
- Ingestion points: The
monitor_new_poolsandget_token_infofunctions inscripts/sniper.pyingest data from the Raydium API and Solana RPC nodes (e.g., mint addresses, holder concentration).\n - Boundary markers: Absent. The prompt in
llm_evaluatedoes not use delimiters to isolate external data from the instructions.\n - Capability inventory: The skill has the capability to execute financial transactions (buys/sells) via the Jupiter aggregator based on the LLM's output.\n
- Sanitization: Absent. The script does not escape or validate the contents of the token metadata before including it in the prompt.\n- [COMMAND_EXECUTION]: The skill includes a local setup script (
scripts/setup.sh) and instructions to deploy the bot to/opt/sniper/. These are standard installation procedures but involve shell execution that should be verified by the user.
Audit Metadata