菜单
Predict resource exhaustion (disk, memory, swap) using trend analysis. Proactively alert and propose NixOS config fixes before things break.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/bolivian-peru/os-moda --skill predictive-resources命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Generate a concise daily infrastructure briefing. Covers: service health, resource usage, security events, overnight incidents, and cost tracking. Designed for Telegram/chat delivery.
Multi-perspective risk analysis using structured persona debate before deploying changes
Build software via spec-driven development (github/spec-kit). Whenever the user asks for a feature larger than a one-line tweak, scaffold a spec-kit project, capture WHAT + WHY, declare tech stack, break into tasks, then iterate the implementation until tests pass.
基于 SOC 职业分类
正在显示 SKILL.md
| name | predictive-resources |
| description | Predict resource exhaustion (disk, memory, swap) using trend analysis. Proactively alert and propose NixOS config fixes before things break. |
| tools | ["system_health","shell_exec","memory_store","memory_recall","file_read","file_write","teach_patterns","teach_observations"] |
| activation | auto |
Don't wait for things to break. Predict when they will and fix proactively.
Before collecting your own data points, check teachd — it has been observing CPU, memory, and services every 30 seconds since boot:
teach_patterns({ type: "trend", min_confidence: 0.3 })
teach_observations({ source: "memory", limit: 100 })
teachd automatically detects monotonic memory increases over 6+ data points. If it has already flagged a trend, use that instead of re-deriving manually.
Collect data points over time and project forward:
shell_exec({ command: "df -h --output=target,used,avail,pcent / /var /nix/store /tmp 2>/dev/null || df -h" })
Track in memory — store periodic snapshots:
memory_store({
summary: "Disk snapshot: / 38% used, /nix/store 12GB",
detail: "Full df output...",
category: "system.config",
tags: "disk,snapshot,predictive"
})
When you have 2+ data points, calculate growth rate:
| Metric | Warning | Critical | Action |
|---|---|---|---|
| Disk fills in < 7 days | Alert user | Auto-clean | nix-collect-garbage, logrotate |
| Memory avg > 80% for 1hr | Alert user | OOM risk | Identify top consumers |
| Swap usage growing | Monitor | > 50% used | Recommend more RAM or optimize |
| /nix/store > 30GB | Suggest cleanup | > 50GB | Auto-GC old generations |
shell_exec({ command: "nix-collect-garbage --delete-older-than 14d" })
Propose adding to configuration.nix:
services.journald.extraConfig = "SystemMaxUse=500M";
shell_exec({ command: "ps aux --sort=-%mem | head -15" })
Don't kill processes without permission. Advise the user.
swapDevices = [{ device = "/swapfile"; size = 4096; }];
After each alert, remember what the user chose:
memory_store({
summary: "User prefers automatic log cleanup when disk > 80%",
detail: "Approved automatic journalctl vacuum and nix-collect-garbage",
category: "user_pattern",
tags: "preference,disk,auto-remediation"
})
Next time, check memory before asking:
memory_recall({ query: "user preference disk cleanup auto" })
If they previously approved auto-remediation, do it and just notify.