AI Coding Agent Collaboration Standard

💡 Copyable AI Prompt Block (AGENTS.md)

<!-- START AGENT-STANDARD: GENERAL-AI-COLLABORATION -->
## AI Coding Agent Collaboration Rules
- Read `CONTEXT.md` and relevant docs before editing code; never guess implementation details or file paths.
- Require explicit user approval before executing file edits, commits, or destructive commands.
- Keep responses concise, evidence-based, and format every file link with the `file://` scheme.
<!-- END AGENT-STANDARD: GENERAL-AI-COLLABORATION -->

---

Detailed Human Guide & Rationale

(Detailed guide to be authored collaboratively)

1. AGENTS.md Rule Enforcement & Block Placement

(Guidelines for structuring system instructions, skill definitions, and project-level rules)

2. Approval Workflows & Human-in-the-Loop Safeguards

(Ensuring AI agents request explicit confirmation for major design decisions, commits, and file edits)

3. Context Engineering & Tool Safety Bounds

(Context window management, targeted search tools, log inspection discipline, and zero-hallucination verification)

---

Primary Evidence & References

• Antigravity AI Agent Rules Specification: System prompt & safety guidelines for agentic pair programming.
• Google Engineering Practices Code Review Guide: https://google.github.io/eng-practices/