scaffold-worker

Warn

Audited by Gen Agent Trust Hub on Jun 18, 2026

Risk Level: MEDIUMDATA_EXFILTRATIONCOMMAND_EXECUTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill instructs the agent to read from an absolute local path: /Users/happypeet/Documents/Github/cc-switch/. This constitutes unauthorized file access/exposure as it targets directories outside the current working directory.
  • [COMMAND_EXECUTION]: The skill executes multiple commands that fetch and run external code, including pnpm create tauri-app and pnpm install. While expected for scaffolding, these commands run arbitrary external scripts at runtime.
  • [INDIRECT_PROMPT_INJECTION]: The work procedure relies on reading external data from the cc-switch project to determine configuration steps.
  • Ingestion points: Files located at /Users/happypeet/Documents/Github/cc-switch/ (referenced in SKILL.md).
  • Boundary markers: Absent. There are no instructions to ignore malicious content within those files.
  • Capability inventory: The skill has broad capabilities including file writing, network package installation (pnpm), and shell command execution (pnpm tauri dev, cargo check) across the SKILL.md file.
  • Sanitization: Absent. Data read from the external project is used directly to configure the new project.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jun 18, 2026, 03:47 PM
Security Audit — agent-trust-hub — scaffold-worker