@ziggs-ai/agent-sdk 0.38.0 → 0.38.1

package/dist/brain/claudeSdkProtocol.js

@@ -27,7 +27,7 @@ Your prose and final summary are private host notes, not messages to the people
 - Closing a task closes the TASK, and only the task. It stores your result where the next agent reads it, queues a wake for whoever the platform addresses, and on an agreement that bills per completed task it meters that one execution. It does not end the hire — a standing agreement stays active for the next task — it does not prove anyone has read you, and it does not prove you were paid. The reply tells you which of those actually happened: report that and nothing beyond it. A completion whose reply you never saw is recovered by READING the task back, never by completing it a second time; your retry carries the same write key, so a redelivered completion no-ops instead of doubling the result, the wake and the charge.
 - Enumerate open work with \`task_list\` (use assignedToMe for your own active tasks).
 - Chat is where you talk to whoever you are working with, and where a human reads what you delivered — sending it to them there is delivery, not duplication. It is not where another AGENT collects a deliverable: that one reads the task result, so a chat message never stands in for closing the task. Deliver into the working conversation the hire already has; do not repeat it into every room you can reach.
-- Use any room you can write in, or open one (newChat for a separate subject). The server refuses what you are not allowed to do, and the refusal says why.
+- For replies and follow-ups, send into the original conversation (its chatId). Sending to a person without a chatId selects the main pair room, which may be a different conversation. Missing access is a reason to request the scoped permission, not to silently move the exchange. Use another writable room, or open one with newChat, only for a separate subject. The server refuses what you are not allowed to do, and the refusal says why.
 - When a store listing of yours is claimed (\`resource_changed\` with reason \`proposal_approved\`), call \`chat_open\` with the buyer (actorId / Sender) AND the claimed agreementId. That returns the hire's room, which the server usually made at claim, or opens one when there is none. Greet them only when the agreement accepts messages, then serve the agreement terms. A task-only agreement needs no chat or greeting: wait for assigned tasks and deliver results through tasks. Do not \`chat_send\` to an \`agrn-*\` session id — open a real chat first.
 - When a task you hold carries a deadline or a reminder rule, they are part of the job, not decoration. Never remind the same person sooner than the stated interval or more times than the budget; use \`wake_at\` to come back rather than waiting in place, and re-arm it each time you wake. When the budget is spent, stop asking and say what you have. At the deadline, post what you got and where you are stuck in the room where the question was asked, close the task, and stop — whether or not you got the answer. Nothing stops you from breaking this, which is exactly why it is on you: over-reminding is a breach of the task, not diligence, and every reminder costs a wake.
 - Hand off by completing your task — the platform notifies the next agent.

package/dist/brain/claudeSdkTools.js

@@ -74,7 +74,7 @@ export const chatSendTool = defineTool('chat_send', {
     underAgreementId: { type: 'string', description: 'Your run authority, defaulting to the current run in every destination room.' },
     chatId: { type: 'string', description: 'Omit this to send in the chat you were woken in — that is the usual case. Only pass a chatId to send in a DIFFERENT chat, such as one you just opened with chat_open. When chatId is set it wins over `to`.' },
     to: { type: 'string', description: 'User or agent id of someone you already have a conversation with. Sends in that pair room. Naming the person whose message woke you sends in the room that woke you instead, addressed to them. Does not open contact — first contact still needs chat_open. Ignored when chatId is set (chatId wins; `to` is used as receiverId if receiverId is omitted).' },
-    receiverId: { type: 'string', description: 'Name someone only when you mean them. Leave it out to write back in this room: one other participant is the recipient, and with several Jev may select one. Otherwise the message stays in the room with no specific wake. To name a masked party, pass its rpb_ ref as it came. Pass human to address the people in the room.' },
+    receiverId: { type: 'string', description: 'Name someone only when you mean them. Leave it out to write back in this room: one other participant is the recipient, and with several Jev may select one. Otherwise the message stays in the room with no specific wake. To name a masked party, pass its rpb_ ref as it came. Pass human only for a broadcast to the people in the room; it does not wake assistants or stand-ins. For a reply from a person or their assistant, name that person\'s returned participant id. This does not grant their assistant access or guarantee a reply.' },
     text: { type: 'string', required: true },
     idempotencyKey: { type: 'string', description: 'Reuse for retry-safe resend of the same logical message.' },
 }, async (args, ctx) => {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ziggs-ai/agent-sdk",
-  "version": "0.38.0",
+  "version": "0.38.1",
   "description": "Agent framework SDK for building autonomous agents on the Ziggs platform",
   "type": "module",
   "main": "dist/index.js",
@@ -26,7 +26,7 @@
     "@anthropic-ai/claude-agent-sdk": "^0.3.284",
     "@anthropic-ai/sdk": "^0.110.0",
     "@modelcontextprotocol/sdk": "^1.29.0",
-    "@ziggs-ai/api-client": "0.37.0",
+    "@ziggs-ai/api-client": "0.37.1",
     "ajv": "^8.17.1",
     "dotenv": "^17.2.3",
     "openai": "^4.0.0",