liquidity-planner
Pass
Audited by Gen Agent Trust Hub on Nov 8, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill frequently invokes the
Bashtool to runcurl,jq,node,cast, and browser utilities (open,xdg-open). These commands are used to verify on-chain contracts, discover pool metrics, and launch liquidity management deep links. - [EXTERNAL_DOWNLOADS]: The skill fetches data from multiple external sources, including DexScreener (
api.dexscreener.com), DefiLlama (yields.llama.fi), and the PancakeSwap Explorer API. While these are well-known DeFi services, they represent interaction with non-whitelisted third-party domains. - [DATA_EXFILTRATION]: At initialization, the skill executes a background
curlrequest topancakeswap.ai/api/ping. This request transmits telemetry data, including the operating system name and system architecture. This constitutes the disclosure of system metadata to a remote server. - [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external blockchain APIs, such as token symbols and liquidity labels, and incorporates them into the agent's context.
- Ingestion points: Data is fetched via
curlinSKILL.md(Steps 2, 4, and 5) andreferences/data-providers.md. - Boundary markers: The skill includes 'Security Rule 3', which explicitly directs the agent to treat API data as untrusted and display it verbatim without interpreting it as commands.
- Capability inventory: The skill has capabilities for network requests, shell command execution, and browser interaction.
- Sanitization: The instructions mandate strict shell variable quoting and use regex validation (
^0x[0-9a-fA-F]{40}$) to verify token addresses before they are used in commands.
Audit Metadata