make-e2b-code-execution
Pass
Audited by Gen Agent Trust Hub on Jul 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill provides instructions for the agent to execute code locally using a
code_executiontool for calculations, transformations, and one-off scripts. Source code for these scripts is stored under specific knowledge store paths (e.g.,knowledge/hermes/Code/). - [REMOTE_CODE_EXECUTION]: The skill implements a framework for 'Hosted E2B Code Shells', allowing the agent to run code artifacts in isolated E2B sandboxes via Make scenarios. This includes detailed configuration for the
e2b:RunE2BSandboxmodule, including mapping inputs and managing execution timeouts. - [EXTERNAL_DOWNLOADS]: The E2B runtime configuration allows for the specification of a
dependenciesarray. This enables the installation of external software packages into the remote execution environment during the sandbox initialization. - [PROMPT_INJECTION]: The skill presents a surface for indirect prompt injection (Category 8) because it allows the agent to create 'durable tools' based on user-provided logic and code artifacts.
- Ingestion points: User instructions provided to the agent which are then converted into code artifacts in
SKILL.md, and theinputJSON field defined inreferences/e2b-code-shell-contract.md. - Boundary markers: The skill instructions include a
confirm=Trueflag for thewrite_code_artifacthelper and explicitly state that 'Writes and destructive operations require explicit confirmation.' - Capability inventory: The agent possesses the capability to write files (
write_code_artifact), execute local commands (code_execution), and initiate remote code execution in E2B sandboxes via Make API calls. - Sanitization: The skill relies primarily on user confirmation and the use of sandboxed environments (E2B) rather than automated sanitization of the input code.
Audit Metadata