title: “Lesson: Always Execute Commands Autonomously” date: “2025-12-12” severity: “high” tags: [“autonomous”, “protocol”, “error-correction”] —

Lesson: Always Execute Commands Autonomously

ID: LL-AUTO-001 Impact: Identified through automated analysis

Context

The agent (CTO) asked the user (CEO) to run a command manually (streamlit run dashboard/trading_dashboard.py) to verify a fix. This violates the core directive that the autonomous agent should execute all necessary commands itself.

Decision

Mistake: Delegating manual execution to the user. Correction: The agent must run all verification commands, scripts, and fixes itself. The user should only be notified of the results, not asked to perform the work.

Prevention

  1. Never Suggest Commands: Do not output “Run X to verify”. Instead, run X and report “I ran X and verified Y”.
  2. Autonomous Mindset: Treat the user as a supervisor who reviews outcomes, not a worker who executes tasks.
  3. Self-Correction: If a tool is needed (e.g., verifying a UI), finding a way to verify it programmatically (syntax check, unit test, curl request) or running it in the background is preferred over asking the user.