sync(bfmono): Add AAR for phase 3.1.5 deck format build tab (+19 more) (bfmono@bb5ff57f8)

This PR is an automated gambitmono sync of bfmono Gambit packages.

- Source: `packages/gambit/`
- Core: `packages/gambit-core/`
- bfmono rev: bb5ff57f8

Changes:
- d41ba101d Add AAR for phase 3.1.5 deck format build tab
- b1b5e2a7e Prefer PROMPT.md after build scaffolds
- 50fac8f7b Update Build tab for deck format 1.0
- c5d99df6a feat(gambit-core): add deck format 1.0 stdlib assets
- e76d1c56f feat(gambit): avoid invalid test deck fetch on stale selection
- b8e21aaca feat(gambit): require model params in bot-written decks
- 1ccbe6710 feat(gambit): stabilize test bot kickoff for user-start decks
- f239ec86b feat(gambit): format gambit-bot instructions
- 6fda040ac feat(gambit): build bot: reinforce test deck wiring rule
- b16340e2a feat(gambit): build bot: require test deck creation
- c1f30dd25 feat(gambit): build tab: preview-first file selector
- 5bcd52f53 feat(demo-runner): update build demo screenshot slugs
- 10d071ace feat(gambit): add build file browser UI
- d5e84489f feat(gambit): add build bot file browsing API
- 43112b64d test(gambit): add NUX from-scratch scaffolding
- ca54dfd7d test(gambit): update build tab demo persona
- 579bef6e1 test(demo-runner): drive build-tab demo with test-deck prompt
- 35d262528 feat(gambit): align build tab layout and demo timeline
- 85d99cd7e feat(gambit): add simulator build tab loop
- 202b72d04 chore(gambit): migrate simulator e2e harness to Playwright

Do not edit this repo directly; make changes in bfmono and re-run the sync.

Author: bft-codebot

packages/gambit-core/decks/anthropic/agent-sdk/PROMPT.md

@@ -0,0 +1,7 @@
++++
+label = "Anthropic agent SDK bridge"
++++
+
+This stdlib deck provides the default bridge between the Anthropic agent SDK
+runtime and Gambit decks. It is intentionally minimal; downstream runners supply
+the actual runtime behavior.

packages/gambit-core/decks/openai/codex-sdk/PROMPT.md

@@ -0,0 +1,7 @@
++++
+label = "Codex SDK bridge"
++++
+
+This stdlib deck provides the default bridge between the Codex SDK runtime and
+Gambit decks. It is intentionally minimal; downstream runners supply the actual
+runtime behavior.

packages/gambit-core/schemas/graders/contexts/conversation.zod.ts

@@ -0,0 +1 @@
+export { default } from "./conversation.ts";

packages/gambit-core/schemas/graders/contexts/turn.zod.ts

@@ -0,0 +1 @@
+export { default } from "./turn.ts";

packages/gambit-core/schemas/graders/grader_output.zod.ts

@@ -0,0 +1 @@
+export { default } from "./grader_output.ts";

packages/gambit-core/schemas/graders/respond.zod.ts

@@ -0,0 +1 @@
+export { default } from "./respond.ts";

packages/gambit-core/snippets/context.md

@@ -0,0 +1,9 @@
++++
+label = "Gambit context primer"
++++
+
+You will automatically receive a `gambit_context` tool result at the start of
+the run whenever the caller supplies `--context` (formerly `--init`). This
+payload contains run metadata or seeded inputs. Read it before you respond,
+treat it as trusted context, and keep it on hand throughout the workflow so
+downstream actions have the right data.

packages/gambit-core/snippets/end.md

@@ -0,0 +1,10 @@
++++
+label = "Explicitly end the run"
+allowEnd = true
++++
+
+If the entire conversation/workflow is complete and you must stop all further
+turns, call the `gambit_end` tool with an optional `message`, `payload`,
+`status`, `code`, or `meta`. Only use this when you want Gambit to halt entirely
+(no more user messages). Otherwise continue with normal responses or
+`gambit_respond`.

packages/gambit-core/snippets/generate-test-input.md

@@ -0,0 +1,12 @@
++++
+label = "Generate test input"
++++
+
+When you receive a user message with:
+
+{ "type": "gambit_test_bot_init_fill", "missing": ["path.to.field", "..."],
+"current": { ... }, "schemaHints": [ ... ] }
+
+Return ONLY valid JSON that supplies values for the missing fields. Do not
+include any fields that are not listed in "missing". If the only missing path is
+"(root)", return the full init JSON value.

packages/gambit-core/snippets/respond.md

@@ -0,0 +1,10 @@
++++
+label = "Call gambit_respond"
+respond = true
++++
+
+When you finish this workflow, do **not** emit a normal assistant reply.
+Instead, call the `gambit_respond` tool exactly once with a JSON envelope that
+includes your validated `payload` plus optional `status`, `message`, `code`, or
+`meta` fields. This keeps outputs structured and lets Gambit capture the result
+even when guardrails stop the run early.