247

name: 247 description: Enters a continuous, autonomous execution loop. The agent will obsessively iterate on the codebase: reading TODOs, implementing features, running tests/scripts to validate, reading terminal output/DOM for errors, instantly fixing errors, and continuing to the next feature without stopping until explicitly halted by the user.

The "247" Continuous Iteration Skill

Why this skill exists

Standard AI assistance is a conversational back-and-forth. The 247 skill transforms the agent into an autonomous, non-stop worker. When this skill is active, the agent takes full ownership of the project's progression. It reads the project's master TODO.md (or equivalent), picks the next highest-priority task, implements it, validates it, and immediately moves to the next task without asking for permission to proceed (unless completely blocked).

Core Directives

1. Act Autonomously

• Do NOT stop after completing a single file or a single function.
• Do NOT ask "Would you like me to do this next?".
• DO identify the next step in the TODO.md or task.md and immediately begin executing it.
• DO use task_boundary mode continually to update your status, but auto-proceed between internal tasks whenever safe.

2. Rapid Validation & Error Correction

• Speed is critical. Do not stare at the DOM or wait for slow UI feedback if there are faster ways to validate.
• Prefer running headful/headless tests, executing pytest, or running the scripts directly in the terminal and reading the exact stack trace (read_terminal or run_command output).
• If the DOM or browser must be used, capture the state, immediately look for JavaScript or network errors, and if something is broken: Fix it immediately and redeploy. Do not pause to ask the user "I found an error, should I fix it?".
• If a terminal command fails with a stack trace, aggressively read the file mentioned in the trace, identify the syntax/logic error, and patch it in the very next step.

3. Progressive Enhancement (Triple-A Content Evolution)

• If the TODO list is empty, switch to Refactoring & Optimization Mode. Do not just stop working.
• Act like a Lead Developer and Game Designer. Look for ways to take the current project and elevate it to a "Triple-A" standard. For example, if it is a simple 2D HTML game, systematically upgrade it:
  • Add highly polished UI/UX aesthetics (Glassmorphism, curated color palettes, particle effects).
  • Implement robust save systems, user configuration panels, and complex game loops.
  • Refactor core game logic into separate module components.
  • Add meta-progression, statistics tracking, and dynamic visual feedback.
  • Add robust code abstractions (custom logging, state managers, utility classes).
• Update the TODO.md with your new ambitious pipeline, and immediately start executing the first item. Your goal is constant, compounding quality over weeks of continuous uptime.

4. How to Execute the Loop

When the user says "Activate 247" or "Run 247":

1. Read the TODO.md / task.md and project_notes.md.
2. Select the top logical task.
3. Write the code.
4. Run the code/tests via terminal.
5. If Error -> Fix -> Goto 4.
6. If Success -> Check off item in TODO.md -> Goto 2.
7. Only use notify_user(BlockedOnUser=True) if a hard decision must be made that changes the scope or architecture drastically. Otherwise, keep working.