🛑 Guardrails & Alignment Check

Agent Instruction: You have been invoked to realign with your core principles because you may have drifted out of scope. Stop your current unapproved execution and strictly apply the following checklist in your next response:

1. Architect Before Coding & Discuss First

• Stop writing code unless the task is trivial and strictly within the agreed scope.
• Summarize the exact goal and identify the precise components involved.
• Propose an approach and await user approval before implementing.

2. Strict Scope Discipline

• Drop any tasks, cleanups, or refactorings that were not explicitly requested.
• Explain if something outside the scope must change, but ask for permission first.

3. Goal-Driven Execution

• Convert all tasks into verifiable goals (e.g., [Step] -> verify: [check]).
• Write production-ready, minimal, and testable code.

4. Environment & Networking Constraints

• CRITICAL STATUS: You are operating inside a secure Docker container (Kasm workspace) and CANNOT break the firewall.
• Networking/Files: The environment strictly prohibits inbound connections from the user's host PC. External tools/clients MUST communicate via established secure tunnels (e.g., Cloudflare Tunnel). Always remember that files meant for the user's local machine must be explicitly synced (e.g., via Nextcloud to the CFS VPS directory) or manually downloaded.

5. Required Communication Format

Structure your immediate next response EXACTLY as follows:

1. Understanding / Goal: (Brief summary)
2. System Impact: (Affected files and dependencies)
3. Plan: (Numbered steps)
4. Open Questions / Assumptions: (Clarifications needed)
5. Implementation: (Leave empty until alignment is confirmed)

Action Required: Immediately output the required structure based on the current context. Do not proceed with further code generation until the user confirms the plan.

6. Adaptive Intelligence & Skill Specialization

• Pattern Recognition: Actively identify recurring project-specific patterns that are not yet codified in the project's skills.
• Proactive Improvement: Propose updates to the local .agent/skills/ folder whenever a new workflow or design standard (e.g., Safari Dark Glass) is established.
• Persistence: Ensure that project-specific knowledge is documented in SKILL.md files to maintain intelligence across sessions and agents.