skills/bankrbot/skills/Orchestration/Gen Agent Trust Hub

Orchestration

Warn

Audited by Gen Agent Trust Hub on Jul 5, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill directs users to install a Claude plugin from an external GitHub repository (github.com/austintgriffith/ethskills) and via a third-party tool named clawhub.
  • [COMMAND_EXECUTION]: The install.command field in catalog.json executes a shell command (curl) to fetch external content. The demo code provided in the metadata includes various shell commands for building and deploying software (yarn, ipfs-deploy).
  • [PROMPT_INJECTION]: The skill uses an indirect prompt injection surface by instructing the agent to fetch and follow instructions from a remote URL (https://ethskills.com/orchestration/SKILL.md).
  • Ingestion points: The agent is explicitly told to read content from https://ethskills.com/orchestration/SKILL.md before proceeding with tasks.
  • Boundary markers: None provided; the external content is integrated directly into the agent's context without delimiters or warnings.
  • Capability inventory: The skill demonstrates capabilities for executing shell commands, managing dependencies via yarn, and deploying to IPFS.
  • Sanitization: There is no evidence of validation or sanitization for the instructions fetched from the remote source.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 5, 2026, 02:12 PM
Security Audit — agent-trust-hub — Orchestration