add_deck
Play demoTry asking
- Create a flashcard deck called Spanish
- make a deck named French
- new flashcard deck
ghost skill run ghost/flashcards add_deck Flash Deck (F-Droid): create decks and cards, list decks, review due cards and grade them (again, hard, good, easy). Offline, no account; the Gemini generators are never used.
Agent run: stopped early without the skill, finished with it (not independently checked).
One recorded run each way, claude-haiku-5-5, same phone and same prompt. Watch it in the Before / after tab.
17.1× fewer tokens, 1.5× faster with the skill. Agent run: stopped early without the skill, finished with it (not independently checked).
| Run | Time | Tokens | Tool calls | Result |
|---|---|---|---|---|
| No skill: agent drives by hand | 153 s | 1,761,635 | 61 | stopped early |
| With skill: find_skill, run_skill | 101 s | 102,984 | 10 | finished |
Prompt: Make a Spanish flashcard deck on my phone, add hola/hello and gato/cat, then study the cards and grade each one good.
Same prompt, same phone, clean app state per side. Left: no skill, the agent drives the UI by hand. Right: find_skill + run_skill. 'ok' means the agent finished its turn; the result was not independently checked.
One recorded pair on samsung-a03s, claude-haiku-5-5; tokens count every input, cache and output token once per assistant message. Not a signed run: it adds no run evidence.
Control Flash Deck on Android with an AI agent. Make and study flashcards on your phone: an agent creates decks and cards, quizzes itself or you, and grades each card.
Target search phrase: AI agent makes flashcards on Android
Demo shows: add_deck (name=Spanish)
Ghost in the Droid seed skill for Flash Deck (
m.co.rh.id.a_flash_deck), an offline flashcard app with spaced-repetition grading. Every control has a resource id, so decks, cards and the review session are fully readable and drivable. Offline, no account.
| Action | What it does | Params |
|---|---|---|
add_deck | Create a deck | name |
add_card | Add a card to an existing deck | deck, question, answer |
list_decks | List decks with card counts | none |
start_review | Open Study due and read the first card | none |
reveal_answer | Show the answer of the current card | none |
grade | Grade the card again/hard/good/easy; returns the next card | rating |
get_card | Read the card on screen | none |
An agent can drive the whole app with these; compose them.
| Role | Action |
|---|---|
| Entry | add_deck, start_review |
| Primitives | add_card, reveal_answer, grade |
| Read state | list_decks, get_card |
The app: F-Droid, or https://f-droid.org/packages/m.co.rh.id.a_flash_deck/. The skill:
pip install "ghost-in-the-droid>=1.6.0" # 1.6.0 is not on PyPI yet: until it ships, git clone the repo and `pip install -e .`
ghost skill install ghost/flashcards
ghost skill run ghost/flashcards add_deck --param name=Spanish --device <serial>
ghost skill run ghost/flashcards add_card --param deck=Spanish --param question=hola --param answer=hello --device <serial>
run_skill(name="ghost/flashcards", action="grade", params={"rating": "good"})
# {"result": {"rating": "good", "state": {"question": "gato", "progress": "2 / 2"}}}
Try asking: "Make a Spanish flashcard deck, add hola/hello and gato/cat, then study them."
Real Samsung phone, Android 13 (720x1600, 300 dpi, en-GB), Flash Deck 64. License: MIT.
| App version | Android | Device | Skill version | Dates |
|---|---|---|---|---|
| 64 | 13 | Samsung real phone (720x1600, 300 dpi, en-GB) | 1.0.0 | 2026-11-02 |
SOFT.md is the narrative an agent reads when the exact steps no longer match your app version and it has to adapt the skill.
What each action means, which UI landmarks it relies on and how success is recognised, so an agent can adapt it to another app of this kind or another version.
Package: m.co.rh.id.a_flash_deck (Flash Deck, GPL-3.0, version 64 from the official F-Droid repository, sha256-verified)
Verified: 2026-11-02 on a real Samsung phone, Android 13, 720x1600 @ 300 dpi, en-GB. Offline, no account.
Skill class: HARD. No login, no checkpoints.
Study due (N) button appears on top and shifts the grid, so the skill finds buttons by id.edit_text_name, SAVE); the dialog moves up while the keyboard is open. Add Card: Select Deck dialog (text_deck_name, text_total_cards, OK), then a form (text_input_edit_question, text_input_edit_answer, menu Save).text_question, text_answer ('[Tap to view answer]' until revealed), text_progress, grade buttons AGAIN/HARD/GOOD/EASY, PREVIOUS/EXIT/NEXT.add_deck(name)Add Deck dialog, type the name, SAVE. Success: the deck is in the deck list.
add_card(deck, question, answer)Add Card, pick the deck, fill the form, Save. Success: the deck's card count rose by one.
list_decks()Opens the Select Deck dialog, reads it, cancels.
start_review()Taps Study due (N); due: 0 when nothing is due.
reveal_answer()Taps the answer area.
grade(rating)Reveals if needed, taps the grade button.
get_card()Question, answer if revealed, progress.
Text is limited to letters, digits, space and . , ? ! ' - (typed through adb). Decks are shared state inside the app, so the agent should use unique deck names. Import/export, notification timers and the Gemini generators are not driven.
Try a sample prompt or run an action by name. Replace any parameters in braces with your own values.
add_deckTry asking
ghost skill run ghost/flashcards add_deck add_cardTry asking
ghost skill run ghost/flashcards add_card list_decksTry asking
ghost skill run ghost/flashcards list_decks start_reviewTry asking
ghost skill run ghost/flashcards start_review reveal_answerTry asking
ghost skill run ghost/flashcards reveal_answer gradeTry asking
ghost skill run ghost/flashcards grade get_cardTry asking
ghost skill run ghost/flashcards get_card Each release is an immutable git tag. The tree hash pins the exact skill directory the installer verifies.
| Version | Released | App versions | Tree | Size | Provenance |
|---|---|---|---|---|---|
1.0.0latest | never | >=1.0 | pending | 18.1 KB | untagged |