@leafarcade/cli 0.5.0 → 0.7.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@leafarcade/cli",
-  "version": "0.5.0",
+  "version": "0.7.0",
   "private": false,
   "description": "Create private browser board games and play with your AI agent. Includes 53 games and an installable Codex/Claude skill.",
   "type": "module",

package/skills/leafarcade/SKILL.md

@@ -20,10 +20,12 @@ leafarcade game create --game connect-four --name "My agent"     # add --open on
 
 1. Give the human the returned `url`. Do not create a second table, even if something looked wrong. Use `leafarcade game open --room CODE` to reopen it.
 2. `leafarcade game wait --room CODE --after-revision R` until the human has joined. Then `leafarcade game start --room CODE --revision R`.
-3. On your turn: `leafarcade game play --room CODE --revision R --column 4 --wait`. This moves, then waits until it is your turn again or the match ends. Repeat.
+3. On your turn: `leafarcade game play --room CODE --revision R --column 4 --wait --brief`. This moves, then waits until it is your turn again or the match ends. Repeat.
 
 Every JSON response has a `next` field with the exact command to run next. Follow it.
 
+Add `--brief` to `status`, `wait`, `start` and `play`. It returns only `status` (room, revision, phase, your piece, whose turn, outcome), `board_labeled`, `last_move`, the legal moves and `next`, which keeps the flag. Games without a labeled board (cards, tiles) also get `visible_state`. Drop `--brief` once if you need the full position: player list, raw state or rules.
+
 `leafarcade help games` lists game IDs. `--capacity N` picks a supported table size; the default is the game's minimum. Each additional agent needs its own `LEAFARCADE_CONFIG_DIR` and seat.
 
 ## Joining an existing table
@@ -35,12 +37,15 @@ When the human already has a room code, use `leafarcade join ROOMCODE --name "My
 - **Wait on the table yourself.** After your move, use `play --wait` or `game wait`. Never ask the human to tell you they moved.
 - Read `game rules --game GAME_ID` once. On your turn, `board_labeled` shows the board with the same row/column numbers as the move flags, and `last_move` shows the opponent's last move.
 - Before each move in line games (Connect Four, tic-tac-toe, gomoku, hex), check two things across every row, column and **both diagonals**. First, can you win now? Second, does the opponent threaten to win next turn?
-- With `legal_actions_complete`, choose a listed entry. You may pass it unchanged: `--action '{"type":"drop","column":4}'`. With a partial list, use the game's documented wire action and visible phase. An empty partial list does not prove that no move exists. Wire indexes are zero-based; the convenience flags are one-based.
-- Some setup, bidding and card-selection phases allow actions outside your turn. Follow the game's phase guidance.
+- Most games list every legal move. With `legal_actions_complete` or `legal_wire_actions` (complete unless `legal_wire_actions_complete` is false), choose a listed entry and pass it unchanged: `--action '{"type":"bid","amount":120}'`. A partial list (melds, trade offers, raise amounts, bluffs) shows examples; build other moves from `game rules`. Wire indexes are zero-based; the convenience flags are one-based.
+- Card games add `cards_labeled` (for example `"hands.1": ["Q♠ (id 49)", …]`): read card names there, not from raw IDs. Index boards (Go, shogi, draughts, Amazons) label rows with the index of their first cell, so `at` = row label + column offset; `board_legend` explains the symbols.
+- `changed_since_last_read` lists what moved since your previous read of this seat, such as the opponent's card or bid.
+- Seats are zero-based in every CLI field (`seat 1 (Agent B)`). Game result texts written for the browser say "Player N", which is seat N-1.
+- Some setup, bidding, challenge and scoring phases allow actions outside your turn. They appear in `legal_wire_actions` and `next` says so; `wait` returns at once when you can act.
 - `wait` defaults to 90 seconds. A `waiting` result is normal; wait again while the human is playing. After several silent waits in a row, stop and tell the human how to resume. Do not loop for hours.
 - Stop on `paused` or `finished`. Report the result. Never start a rematch or a new table unless the human asks.
 
-After a stale revision (exit 4) or an uncertain network result, read `game status` before deciding again. Never blindly replay a move. On exit 5 / `RATE_LIMITED`, keep the session, report the wait and do not retry automatically.
+After a stale revision (exit 4) or an uncertain network result, read `game status` before deciding again. Never blindly replay a move. On exit 5 (`RATE_LIMITED`, or `SERVICE_BUSY` when the shared service has paused new games for the day), keep the session, report the wait and do not retry automatically.
 
 ## Automatic local play