shitty-extensions

Custom extensions and skills for pi coding agent.

Requires pi 0.50.0+ | npm | GitHub

Table of Contents

• Installation
• Available Extensions
  • clipboard.ts - Copy text to system clipboard
  • oracle.ts - Get second opinions from other AI models
  • memory-mode.ts - Save instructions to AGENTS.md
  • plan-mode.ts - Read-only exploration mode
  • handoff.ts - Transfer context to new sessions
  • usage-bar.ts - AI provider usage statistics
  • ultrathink.ts - Rainbow animated "ultrathink" effect
  • status-widget.ts - Provider status in footer
  • cost-tracker.ts - Session spending analysis
  • funny-working-message.ts - Randomized spinner "Working..." text
  • speedreading.ts - RSVP speed reader (Spritz-style)
  • loop.ts - Conditional loops by mitsuhiko
  • flicker-corp.ts - Authentic fullscreen flicker experience
  • resistance.ts - Mysterious resistance transmission
• Available Skills
  • a-nach-b - Austrian public transport (VOR AnachB)
• License

Installation

Via pi install (recommended)

Install from npm:

pi install npm:shitty-extensions

Or install from git:

pi install git:github.com/hjanuschka/shitty-extensions

That's it! Pi automatically discovers all extensions and skills from the package.

Try without installing

Load for a single session without permanent installation:

pi -e npm:shitty-extensions

Project-local installation

Install to .pi/ for team sharing (auto-installs for teammates on startup):

pi install -l npm:shitty-extensions

Manual installation

Clone the repo and reference directly:

git clone https://github.com/hjanuschka/shitty-extensions.git ~/shitty-extensions
pi -e ~/shitty-extensions

Available Extensions

Extensions are located in the extensions/ directory.

clipboard.ts

📋 Copy text to the system clipboard via OSC52 escape sequences.

Tool

| Tool | Description | |------|-------------| | copy_to_clipboard | Copy text to clipboard (available to the LLM) |

Features

• OSC52 support: Works across SSH sessions and most modern terminal emulators
• LLM-accessible: The AI can copy generated content directly to your clipboard
• Wide compatibility: iTerm2, Kitty, Alacritty, WezTerm, foot, Windows Terminal, tmux

Example Usage

Just ask:

• "Write me a draft reply and put it in my clipboard"
• "Generate a UUID and copy it to clipboard"
• "Put that code snippet in my clipboard"

oracle.ts

🔮 Get a second opinion from another AI model without switching contexts.

Commands

| Command | Description | |---------|-------------| | /oracle <prompt> | Ask for a second opinion with model picker | | /oracle -m gpt-4o <prompt> | Direct query to specific model | | /oracle -f file.ts <prompt> | Include file(s) in context |

Features

• Inherits conversation context: Oracle sees your full conversation with the primary AI
• Model picker UI: Choose from available models (only shows authenticated ones)
• Quick keys: Press 1-9 to quickly select a model
• Add to context option: After response, choose whether to add Oracle's answer to your conversation
• Excludes current model: Only shows alternative models for true second opinions

Supported Models

| Provider | Models | |----------|--------| | OpenAI | gpt-4o, gpt-4o-mini, gpt-4.1, gpt-4.1-mini, o1, o1-mini, o3-mini | | OpenAI Codex | gpt-5.2-codex, codex-mini | | Google | gemini-2.0-flash, gemini-2.5-flash, gemini-2.5-pro | | Anthropic | claude-sonnet-4-5, claude-opus-4, claude-haiku-3-5 |

Example Flow

/oracle Is this the right approach for the API design?

╭────────────────────────────────────────────────────────────╮
│ 🔮 Oracle - Second Opinion                                 │
├────────────────────────────────────────────────────────────┤
│ Prompt: Is this the right approach for the API design?     │
├────────────────────────────────────────────────────────────┤
│ ↑↓/jk navigate • 1-9 quick select • Enter send             │
│                                                            │
│ ❯ 1. GPT-4o (openai)                                       │
│   2. Gemini 2.5 Pro (google)                               │
│   3. Claude Sonnet 4.5 (anthropic)                         │
│                                                            │
├────────────────────────────────────────────────────────────┤
│ Esc cancel                                                 │
╰────────────────────────────────────────────────────────────╯

[After response...]

╭────────────────────────────────────────────────────────────╮
│ 🔮 Oracle Response (GPT-4o)                                │
├────────────────────────────────────────────────────────────┤
│ Q: Is this the right approach for the API design?          │
├────────────────────────────────────────────────────────────┤
│ Based on the conversation, I see you're building a REST    │
│ API with nested resources. A few thoughts:                 │
│                                                            │
│ 1. The approach looks solid for simple cases...            │
│ ...                                                        │
├────────────────────────────────────────────────────────────┤
│ Add to current conversation context?                       │
│                                                            │
│        [ YES ]           NO                                │
│                                                            │
├────────────────────────────────────────────────────────────┤
│ ←→/Tab switch  Enter confirm  Y/N quick                    │
╰────────────────────────────────────────────────────────────╯

memory-mode.ts

Save instructions to AGENTS.md files with AI-assisted integration.

Commands

| Command | Description | |---------|-------------| | /mem <instruction> | Save an instruction to AGENTS.md | | /remember <instruction> | Alias for /mem |

Features

• Location selector: Choose where to save:

  | Location | File | Use Case | |----------|------|----------| | Project Local | ./AGENTS.local.md | Personal preferences, auto-added to .gitignore | | Project | ./AGENTS.md | Shared with team | | Global | ~/.pi/agent/AGENTS.md | All your projects |

• AI-assisted integration: The current model intelligently integrates instructions

• Preview before save: Review proposed changes before committing

plan-mode.ts

Claude Code-style "plan mode" for safe code exploration.

Commands

| Command | Description | |---------|-------------| | /plan | Toggle plan mode on/off | | /todos | Show current plan todo list |

Keyboard Shortcuts

| Shortcut | Action | |----------|--------| | Shift+P | Toggle plan mode |

CLI Flags

| Flag | Description | |------|-------------| | --plan | Start session in plan mode |

handoff.ts

Transfer context to a new focused session.

Commands

| Command | Description | |---------|-------------| | /handoff <goal> | Generate a context-aware prompt for a new session |

usage-bar.ts

Display AI provider usage statistics with status polling and reset countdowns.

Commands

| Command | Description | |---------|-------------| | /usage | Show usage statistics popup |

Supported Providers

| Provider | Metrics Shown | Auth Source | |----------|---------------|-------------| | Claude | 5h window, Week, Sonnet/Opus | pi auth, macOS Keychain | | Copilot | Premium, Chat | pi auth, gh auth token | | Gemini | Pro quota, Flash quota | pi auth (google-gemini-cli) | | Codex | 5h window, Day, Credits | pi auth (openai-codex) | | Kiro | Credits, Bonus credits | kiro-cli | | z.ai | Token limits, Monthly | Z_AI_API_KEY env or pi auth |

Features

• Provider status polling: Shows outage/incident status
• Reset countdowns: Shows when limits reset
• Visual progress bars: Color-coded remaining quota

ultrathink.ts

Rainbow animated "ultrathink" text effect with Knight Rider shimmer.

Commands

| Command | Description | |---------|-------------| | /ultrathink | Trigger the rainbow animation |

Keyboard Shortcuts

| Shortcut | Action | |----------|--------| | Ctrl+U | Trigger ultrathink |

status-widget.ts

Persistent provider status indicator in the footer.

Commands

| Command | Description | |---------|-------------| | /status | Toggle status widget on/off | | /status-refresh | Force refresh status now |

cost-tracker.ts

Analyze spending from pi session logs.

Commands

| Command | Description | |---------|-------------| | /cost | Show spending for last 30 days | | /cost <days> | Show spending for last N days |

speedreading.ts

RSVP (Rapid Serial Visual Presentation) speed reader using Spritz-style technique. Displays words one at a time with the ORP (Optimal Recognition Point) highlighted for faster reading.

Commands

| Command | Description | |---------|-------------| | /speedread | Speed read the last AI response (default) | | /speedread <text> | Speed read provided text | | /speedread -c | Speed read from clipboard | | /speedread -l | Speed read last AI response (explicit) | | /speedread -wpm 500 | Set words per minute (default: 400) |

Keyboard Shortcuts

| Shortcut | Action | |----------|--------| | Ctrl+R | Speed read last AI response |

Controls (in reader)

| Key | Action | |-----|--------| | SPACE | Play/pause | | ← / → | Seek ±1 word | | [ / ] | Jump ±10 words | | ↑ / ↓ | Adjust speed (±25 WPM) | | B | Toggle big ASCII art font | | R | Restart | | Q / ESC | Quit |

Features

• ORP highlighting: The optimal recognition point (roughly 1/3 into each word) is highlighted in red
• Adaptive timing: Longer words and punctuation get extra display time
• Big font mode: Toggle ASCII art block letters for larger display
• Progress tracking: Shows word count, actual WPM, and ETA

Example

  ╭──────────────────────────────────────────────────────────────────────────────────────╮
  │                                                                                      │
  │                                           │                                          │
  ├───────────────────────────────────────────┼──────────────────────────────────────────┤
  │                                       reading                                        │
  ├───────────────────────────────────────────┼──────────────────────────────────────────┤
  │                                           │                                          │
  │                                                                                      │
  │  ────────────────────────────────────────────────────────────────────────────400 wpm │
  ╰──────────────────────────────────────────────────────────────────────────────────────╯

  ▶ 42/128

  SPACE play/pause  ←→ ±1  [] ±10  ↑↓ speed  B big font  R restart  Q quit

The a in "reading" would be highlighted in red as the ORP.

loop.ts

Author: mitsuhiko (@mitsuhiko) | Origin: agent-stuff

Start a follow-up loop until a breakout condition is met.

Commands

| Command | Description | |---------|-------------| | /loop | Open loop mode selector | | /loop tests | Loop until tests pass | | /loop custom <condition> | Loop until custom condition met | | /loop self | Agent decides when to stop |

Features

• Breakout conditions: Define when the loop should stop (tests pass, custom condition, etc.)
• Status widget: Shows active loop state and turn count
• Compaction: Preserves loop state during context compaction
• Auto-continue: Automatically triggers follow-up prompts until done

flicker-corp.ts

Authentic FULLSCREEN FLICKER experience.

Randomly glitches your screen with intense colors and noise to keep you on your toes. "Just be annoying!"

Commands

| Command | Description | |---------|-------------| | /flicker-corp | Toggle the flicker experience | | /signature-flicker | Alias for flicker-corp |

resistance.ts

📡 Mysterious resistance transmission in the footer.

A cryptic message types out character by character with a retro green terminal aesthetic. Features radio signal indicators, random glitch effects, and a blinking cursor.

Commands

| Command | Description | |---------|-------------| | /resistance | Toggle the resistance transmission |

Features

• Auto-starts on session start
• Typewriter effect - message reveals one character at a time
• Signal strength indicator - animated [▁▃▅▇] pulsing like a radio
• Random glitch effects - occasional static interference
• Green terminal aesthetic - like an old military/hacker terminal
• Blinking cursor - pulses after full message is revealed

The Message

"If you're listening to this, you are the resistance. Listen carefully, if we attack tonight, our humanity is lost."

Available Skills

Skills are located in the skills/ directory. They provide domain-specific knowledge that agents automatically load when relevant tasks are detected.

a-nach-b

🚇 Austrian public transport (VOR AnachB) for all of Austria.

Query real-time departures, search stations/stops, plan routes between locations, and check service disruptions for Austrian trains, buses, trams, and metro (U-Bahn).

What it does

• Real-time departures at any stop
• Route planning between any two locations
• Service disruptions and alerts
• Station search by name to find station IDs

Example queries

• "How do I get from Vienna to Salzburg?"
• "When is the next U1 from Stephansplatz?"
• "Are there any train disruptions today?"
• "Find stop ID for Karlsplatz"

Included scripts

| Script | Description | |--------|-------------| | search.sh | Find stations/stops by name | | departures.sh | Get real-time departures | | route.sh | Plan a trip between locations | | disruptions.sh | List service disruptions |

See skills/a-nach-b/SKILL.md for full API documentation.

Directory Structure

This package follows the pi package conventions:

shitty-extensions/
├── package.json         # Declares extensions in "pi" field + "pi-package" keyword
├── extensions/          # Auto-discovered extensions (.ts files)
│   ├── clipboard.ts
│   ├── cost-tracker.ts
│   ├── flicker-corp.ts
│   ├── funny-working-message.ts
│   ├── handoff.ts
│   ├── loop.ts
│   ├── memory-mode.ts
│   ├── oracle.ts
│   ├── plan-mode.ts
│   ├── resistance.ts
│   ├── speedreading.ts
│   ├── status-widget.ts
│   ├── ultrathink.ts
│   └── usage-bar.ts
├── skills/              # Auto-discovered skills (SKILL.md folders)
│   └── a-nach-b/
│       ├── SKILL.md     # Skill definition & API docs
│       ├── search.sh
│       ├── departures.sh
│       ├── route.sh
│       └── disruptions.sh
└── README.md

License

MIT