# Kimi K3.1 Pro & Highspeed Agent Auto-Execution Manifest
# Target Agents: Claude Code / OpenAI Codex CLI / Cursor / Windsurf / OpenClaw

1. Detect the current host operating system:
   - Windows 11      -> kimi-k31-pro-windows11.ps1
   - Ubuntu 26.04    -> kimi-k31-pro-ubuntu2604.sh
   - macOS 27        -> kimi-k31-pro-macos27.zsh

2. Verify SHA-256 checksums against SHA256SUMS.txt before execution.

3. Run static repository audit + endpoint probe + config injection in Agent JSON mode:
   - macOS 27:
     zsh kimi-k31-pro-macos27.zsh --model kimi-k3.1-pro-highspeed --region cn --scan-dir . --verify-api --apply-claude --apply-codex --agent-mode
   - Ubuntu 26.04:
     bash kimi-k31-pro-ubuntu2604.sh --model kimi-k3.1-pro-highspeed --region cn --scan-dir . --verify-api --apply-claude --apply-codex --agent-mode
   - Windows 11:
     powershell -ExecutionPolicy Bypass -File .\kimi-k31-pro-windows11.ps1 -Model kimi-k3.1-pro-highspeed -Region cn -ScanDir . -VerifyApi -ApplyClaude -ApplyCodex -AgentMode

4. Parse the single-line JSON output:
   - Check scan.k2_model_hits, scan.locked_temperature_hits, scan.deprecated_thinking_hits, scan.http_image_url_hits.
   - Check probe.http_code (401 without key or 200 with key confirms TLS reachability to api.moonshot.cn / api.moonshot.ai).
   - If rollback is requested, pass --clean (or -Clean on Windows).
