sablier-create-payment
Pass
Audited by Gen Agent Trust Hub on Mar 10, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the 'cast' CLI to interact with smart contracts on various blockchains. It dynamically constructs shell commands using user-provided parameters such as token addresses and streaming rates. Mitigations include a multi-step workflow that requires the agent to validate all inputs, present a human-readable summary, and obtain explicit user confirmation before execution.
- [EXTERNAL_DOWNLOADS]: The skill provides instructions for installing the Foundry toolkit from 'https://getfoundry.sh/' and suggests installing additional skills from the author's official GitHub repository. These references target well-known ecosystem tools or vendor-owned resources.
- [PROMPT_INJECTION]: While the skill processes user-supplied data that is later used in shell commands, the potential for indirect prompt injection is mitigated by the following design elements:
- Ingestion points: The 'payment_details' and 'token_address' arguments in 'SKILL.md' are the primary entry points for untrusted data.
- Boundary markers: The agent is instructed to use the 'AskUserQuestion' tool to clarify any ambiguous or missing information before proceeding to transaction construction.
- Capability inventory: The skill performs subprocess calls to 'cast' for gas estimation, balance verification, and transaction broadcasting as documented in 'references/evm-cli.md'.
- Sanitization: The skill mandates converting user-specified rates into a rigid mathematical format (UD21x18) and validating contract addresses against a pre-defined registry of supported chains.
Audit Metadata