obelisk-installer

Pass

Audited by Gen Agent Trust Hub on Sep 20, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The project instructions recommend installing the Obelisk CLI by piping a shell script from the author's GitHub repository (raw.githubusercontent.com/tommy0103/obelisk/main/install.sh) to the shell. This is a vendor-owned resource used for the official installation process.
  • [DYNAMIC_EXECUTION]: The skill implements a 'CodeAct' memory layer where the agent writes JavaScript queries to temporary files and executes them via the CLI. This execution is performed within a node:vm sandbox which restricts access to sensitive resources like the filesystem or network.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes historical session transcripts, which are untrusted data. The skill mitigates this risk by using a sandboxed execution environment and relying on the agent's built-in safety guidelines for processing retrieved content.
  • [EXTERNAL_DOWNLOADS]: The skill and its associated tools download dependencies from well-known and trusted registries such as npm (npmjs.com) and utilize resources from official vendor repositories on GitHub.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 20, 2026, 05:21 AM
Security Audit — agent-trust-hub — obelisk-installer