AI Shell — Prompt Your OS

Talk to your operating system in plain English. RobOS ships two tightly-integrated apps — AI Prompt and Skills Manager — that together turn natural language into shell commands, run them with a real AI agent, and return structured step-by-step results.

Table of contents

1. Overview
2. RobOS AI Prompt
   1. What it does
   2. User Interface
   3. Skill selection and parameter inputs
   4. Skills-only mode
   5. AI agent selection
   6. Structured results format
   7. Run history
3. Skills Manager
   1. What it does
   2. Built-in skill library (74 skills)
      1. File Operations
      2. Process Management
      3. Git
      4. Network, Docker, System, Package Management, Text Processing, Security, Development
   3. Custom skills
   4. Community skill packs
4. Architecture
5. IPC Reference
   1. AI Prompt (ai-prompt/main.js)
   2. Skills Manager (skills-manager/main.js)
   3. Skill object schema

---

Overview

Traditional shells require memorizing dozens of command-line tools, flags, and syntax. RobOS AI Shell flips that model:

1. Browse pre-built skills in the sidebar (or search by name)
2. Fill in any required parameters (filenames, ports, patterns) — they appear as inline input fields
3. Type an optional prompt or skip it entirely if the skills say it all
4. Pick your AI agent (Claude, Copilot, Codex, Gemini) via the header pill
5. Hit Run — the AI executes the commands and returns a structured report: each step has the command, its raw output, and a plain-English explanation

Every run is stored in history so you can replay or reference past operations.

---

RobOS AI Prompt

What it does

AI Prompt is a conversational shell. It bridges natural language and the Linux command line by:

• Letting you describe what you want in plain English
• Attaching pre-built shell “skills” (commands) to guide the AI
• Running everything through a real AI agent that handles execution and explanation
• Returning a numbered step report with commands, output, and notes

User Interface

┌─────────────────────────────────────────────────────────────────┐
│ 🤖 RobOS AI Prompt          [Agent: ⊕ Copilot ▾]  [History]   │
├───────────────┬─────────────────────────────────────────────────┤
│ Skills        │                                                 │
│ ┌ Search ──┐  │  Selected skills:                              │
│ │           │  │  ┌────────────────┐  ┌────────────────────┐   │
│ File Ops    │  │  │ disk-usage  ✕ │  │ grep-recursive  ✕ │   │
│  disk-usage │  │  └────────────────┘  │ PATTERN: [______] │   │
│  find-large │  │                      └────────────────────┘   │
│  json-pretty│  │  ┌─────────────────────────────────────────┐  │
│ Git         │  │  │ What do you want to do?                  │  │
│  git-status │  │  │                                          │  │
│  git-log    │  │  └─────────────────────────────────────────┘  │
│  git-stash  │  │                                                 │
│ Process     │  │  [▶ Run with AI]  [Clear]                     │
│  list-procs │  │  Select a skill or enter a prompt              │
│  kill-port  │  │                                                 │
│  top-cpu    │  ├─────────────────────────────────────────────────┤
│  ...        │  │ ✅ Results                                      │
│             │  │  Step 1: df -h --output=size,used,avail,pcent  │
│             │  │  [output]  Home directory is 43% full (23 GB)  │
└─────────────┴─────────────────────────────────────────────────────┘

Skill selection and parameter inputs

When you click a skill in the sidebar it gets added to the selected skills area above the prompt box. Skills that contain $VARIABLE placeholders automatically expand into an input card:

┌──────────────────────────────┐
│  grep-recursive          ✕  │
│  PATTERN: [____________]     │
└──────────────────────────────┘

The parameter value is substituted directly into the command before it reaches the AI. If you leave a parameter blank, the raw $PATTERN token is passed through — the AI will attempt to infer it from your prompt.

Skills-only mode

You don’t need to type a prompt. If you’ve selected at least one skill, the Run button is enabled immediately. The AI receives the auto-prompt:

“Run the selected skills and show me the results.”

Combine multiple skills — for example git-status + disk-usage + list-processes — for a quick system health report with a single click.

AI agent selection

The agent pill in the header lets you switch AI provider mid-session:

Provider	Command used
GitHub Copilot	gh copilot suggest -t shell
Claude	claude -p
Codex	codex
Gemini	gemini

Auth status is checked when the app loads. If a provider isn’t authenticated, a banner appears with a login link.

Structured results format

Every run returns a JSON structure with:

{
  "summary": "Checked disk usage and found 3 large log files",
  "status": "success",
  "steps": [
    {
      "step": 1,
      "action": "Check disk usage",
      "command": "df -h ~",
      "output": "Filesystem  Size  Used  Avail  Use%\n/dev/sda1  100G   43G   57G   43%",
      "note": "Home directory is 43% full with 57 GB remaining."
    }
  ]
}

The renderer displays each step as a card: command in a monospace code block, output collapsible, note in plain prose.

Run history

Every prompt+result pair is stored in ~/.config/robos/ai-prompt-history.json. The history panel (top right) lists past runs newest-first. Click any entry to restore the prompt text and view the original results.

---

Skills Manager

What it does

Skills Manager is the library catalog for AI Prompt. It lets you:

• Browse all built-in skills grouped by category
• Search by name, description, or category
• Create custom skills with your own shell commands and $PARAM definitions
• Import community skill packs from GitHub repositories

Built-in skill library (74 skills)

File Operations

| Skill | Command | Params | |:——|:——–|:——-| | disk-usage | df -h --output=source,size,used,avail,pcent | — | | find-large-files | find ~ -type f -size +100M 2>/dev/null \| head -20 | — | | list-directory | ls -la | — | | json-pretty | cat $FILE \| jq . | $FILE | | csv-summary | head -5 $FILE && wc -l $FILE | $FILE | | count-lines | wc -l $FILE | $FILE | | file-permissions | ls -la $FILE | $FILE | | find-by-extension | find . -name "*.js" \| head -30 | — |

Process Management

| Skill | Command | Params | |:——|:——–|:——-| | list-processes | ps aux --sort=-%cpu \| head -20 | — | | kill-port | fuser -k $PORT/tcp 2>/dev/null \|\| lsof -ti:$PORT \| xargs kill -9 2>/dev/null | $PORT | | top-cpu | ps aux --sort=-%cpu \| head -10 | — | | top-memory | ps aux --sort=-%mem \| head -10 | — | | port-in-use | ss -tlnp \| grep "$PORT" | $PORT | | zombie-processes | ps aux \| awk '$8 == "Z"' | — |

Git

| Skill | Command | Params | |:——|:——–|:——-| | git-status | git status && git log --oneline -10 | — | | git-log | git log --oneline --graph --decorate -20 | — | | git-branches | git branch -a | — | | git-stash | git stash list | — | | git-cleanup | git branch --merged \| grep -v '\*\|main\|master' \| xargs git branch -d | — | | git-remotes | git remote -v | — | | git-diff | git diff --stat HEAD | — | | uncommitted | git status --short | — |

Network, Docker, System, Package Management, Text Processing, Security, Development

(See Skills Manager app for the full list — all 74 skills are browsable and searchable in the sidebar)

Custom skills

Create a custom skill via the New Skill button. Fields:

Field	Description
ID	Unique slug (e.g. deploy-app)
Name	Display name in the sidebar
Category	Groups the skill in the sidebar
Description	One-line tooltip
Command	Shell command. Use $VAR for parameters.

Custom skills are saved to ~/.config/robos/skills.json and immediately available in AI Prompt.

Community skill packs

The Skills Manager can install skill packs from any GitHub repository that exports a JSON array of skill objects.

How to add a pack:

1. Click Add Pack in the sidebar footer
2. Paste the GitHub repository URL (e.g. https://github.com/example/shell-skills)
3. The app clones the repo to ~/.config/robos/skill-packs/<repo-name>/
4. Available skills are previewed — check which ones to install
5. Click Install Selected — they’re merged into ~/.config/robos/skills.json

Required skill pack format — the repo must contain a skills.json at its root:

[
  {
    "id": "my-skill",
    "name": "My Skill",
    "description": "Does something useful",
    "category": "Custom",
    "command": "echo hello $NAME",
    "params": ["NAME"]
  }
]

---

Architecture

packages/
├── skills-manager/
│   ├── main.js           ← Electron main; IPC handlers
│   ├── skills-data.js    ← Shared BUILTIN_SKILLS array (74 skills)
│   └── renderer/
│       ├── index.html
│       ├── app.js
│       └── style.css
│
└── ai-prompt/
    ├── main.js           ← loads skills-data.js + ~/.config/robos/skills.json
    ├── preload.js
    └── renderer/
        ├── index.html
        ├── app.js        ← extractParams(), substituteParams(), renderSkillChips()
        └── style.css

Shared data flow:

skills-manager/skills-data.js
          │
          ├── required by skills-manager/main.js  (sm-list-skills IPC)
          └── required by ai-prompt/main.js        (ap-list-skills IPC)
                                  │
                    ~/.config/robos/skills.json  (custom/community)
                                  │
                        merged → renderer sidebar

skills-data.js is the single source of truth for built-in skills. To add a built-in skill, add one entry to that file — both Skills Manager and AI Prompt will automatically pick it up on next launch.

---

IPC Reference

AI Prompt (ai-prompt/main.js)

Channel	Input	Returns
ap-list-skills	—	{ builtin: Skill[], custom: Skill[] }
ap-run-prompt	{ prompt, skills, agent }	{ ok, summary, steps[], error? }
ap-list-history	—	HistoryEntry[]
ap-save-history	HistoryEntry	{ ok }
robos-check-agent-auth	{ agent }	{ ok, authenticated }

Skills Manager (skills-manager/main.js)

Channel	Input	Returns
sm-list-skills	—	{ builtin: Skill[], custom: Skill[] }
sm-save-skill	Skill	{ ok }
sm-delete-skill	{ id }	{ ok }
sm-import-pack	{ url }	{ ok, skills: Skill[], error? }
sm-install-pack-skills	{ skills: Skill[] }	{ ok, count }

Skill object schema

interface Skill {
  id: string;          // unique slug, e.g. "disk-usage"
  name: string;        // display name, e.g. "Disk Usage"
  description: string; // one-liner for tooltip/search
  category: string;    // sidebar grouping
  command: string;     // shell command, may contain $VAR tokens
  builtin?: boolean;   // true for skills-data.js entries
}