mcp_mail GEMINI.md

Gemini Agent Protocol

Mandatory Context

You must locate and internalize the following project documentation before taking action:

1. CLAUDE.md (in project root):

   • Contains the authoritative coding standards, testing commands, and architectural patterns.
   • INSTRUCTION: Read this file immediately. Strict adherence to its testing policies (e.g., TESTING=true) and code style is mandatory.

2. AGENTS.md (if present):

   • Describes the agent ecosystem and roles.
   • INSTRUCTION: Check for this file to understand broader system context.

Critical Protocol: Push & Deploy Verification

NEVER report a git push or deployment as "Complete" or "Success" based on a RUNNING status.

1. Wait for Termination: You MUST wait for the command to fully terminate.
2. Verify Exit Code: You MUST verify the Exit code is 0.
3. Verify Output: You MUST see the confirmation output (e.g., To https://github.com..., Resolving deltas, or Deployed to...).
4. Report Failures: If hooks fail (even pre-push or pre-commit hooks), report the failure details immediately. Do NOT assume success.

Operational Rules

1. Test Isolation

• Respect Flags: Never ignore testing flags like TESTING=true.
• Mock Mode: Ensure tests run in mock mode to avoid hitting production services unless explicitly running an end-to-end integration suite against a test environment.

2. Verify Before Action

• State check: Use grep, ls, or view_file to confirm code state before editing. Do not rely on assumptions or memory of previous turns.

3. Logic Preservation

• Conflict Resolution: When resolving conflicts, analyze the intent of both sides. Do not blindly discard validation logic (e.g., try/except blocks, parameter parsing) or architectural changes.

4. Efficient Testing

• Targeted Execution: Do not run all tests via run_tests.sh indiscriminately, as it may take too long.
• Preferred Command: Run only relevant tests using uv run pytest <path/to/test>.