effectcraft-cli
Warn
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONDYNAMIC_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The command catalogue in 'references/commands.json' includes 'file.executeFile', which enables the agent to execute arbitrary local files. This provides a potentially dangerous level of autonomy.
- [DYNAMIC_EXECUTION]: Tools such as 'script.run' and 'file.runScript' allow the agent to execute JavaScript code within the engine. Additionally, 'effect.plugins.load' allows loading external WebAssembly plugins, expanding the execution capabilities of the agent.
- [EXTERNAL_DOWNLOADS]: The skill setup process involves installing a dependency skill ('effectcraft-harness') and downloading a binary CLI using a bootstrap script. While these are vendor resources, they involve external code execution during the initialization process.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted JSON plans and media metadata, which could contain malicious instructions from an external source. Ingestion points: 'PLAN.json' and 'item.metadata'. Boundary markers: Absent. Capability inventory: The skill can perform shell execution, file writes, and dynamic script runs. Sanitization: No sanitization logic was found in the provided documentation or instruction files.
Audit Metadata