luxonis-workspace

Pass

Audited by Gen Agent Trust Hub on Sep 3, 2026

Risk Level: SAFEREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill references a shell command to download and execute an installation script for the vendor's oakctl tool. The source URL https://oakctl-releases.luxonis.com/oakctl-installer.sh belongs to the official vendor domain. The skill includes a specific guardrail requiring the agent to ask for user confirmation before recommending or executing this command.
  • [EXTERNAL_DOWNLOADS]: The skill downloads the oakctl installer and documentation from vendor-controlled subdomains (oakctl-releases.luxonis.com and mcp.luxonis.com). These are recognized as legitimate vendor resources for the skill's primary purpose of environment setup.
  • [COMMAND_EXECUTION]: The skill manages environment setup tasks such as creating Python virtual environments and configuring system udev rules. It explicitly instructs the agent to request user permission before performing actions involving sudo, global package installations, or firmware updates.
  • [INDIRECT_PROMPT_INJECTION]: The skill reads existing configuration files such as AGENTS.md and CLAUDE.md to determine the current state of the workspace. This represents a standard surface for indirect prompt injection. The skill mitigates this by using delimited sections for its updates and focusing on bootstrapping tasks.
  • Ingestion points: SKILL.md reads AGENTS.md, CLAUDE.md, and docs/.
  • Boundary markers: Instructions mention adding or updating a 'clearly delimited Luxonis section'.
  • Capability inventory: File writes, virtual environment creation, and shell command suggestions.
  • Sanitization: Not explicitly defined, but the skill is limited to project scaffolding.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 3, 2026, 01:16 AM
Security Audit — agent-trust-hub — luxonis-workspace