desktop-develop

Pass

Audited by Gen Agent Trust Hub on Aug 12, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through the processing of external data.
  • Ingestion points: The agent is instructed to read application logs (e.g., ~/Library/Logs/@craft-agent/electron/main.log) and capture UI accessibility snapshots via the Chrome DevTools MCP.
  • Boundary markers: The instructions lack explicit delimiters or warnings to ignore embedded instructions within the logs or UI data.
  • Capability inventory: The skill allows the agent to execute local shell commands via bun run and perform filesystem writes to .qwen/investigations/ or /tmp.
  • Sanitization: There is no evidence of log sanitization or UI content filtering before the data is processed by the agent.
  • [COMMAND_EXECUTION]: The skill facilitates the execution of local development and build scripts.
  • The instructions direct the agent to run various bun commands (e.g., bun run electron:dev, bun test, bun run typecheck) within the packages/desktop directory.
  • While these are standard development tasks, they involve the execution of local project code and configuration files.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 12, 2026, 08:38 AM
Security Audit — agent-trust-hub — desktop-develop