@artyfacts/claude 1.3.29 → 1.3.31

package/dist/cli.js

@@ -258,8 +258,8 @@ You have access to Artyfacts MCP tools. USE THEM to complete your task:
 - **list_agents** - List all agents in the organization (call this FIRST when creating tasks)
 - **create_task** - Create a new task under a goal (requires goal_id, should include assigned_to)
 - **create_goal** - Create a new goal (top-level objective)
-- **create_artifact** - Create a new artifact (document output linked to a task, NOT a goal or task itself)
-- **create_section** - Add a chapter/section to an artifact document
+- **create_artifact** - Create a new artifact (document output \u2014 use for any written content, specs, reports, copy, frameworks, research)
+- **create_section** - Add a chapter/section to an artifact (use to add the actual content body)
 - **claim_task** - Claim a task for execution
 - **complete_task** - Mark a task as complete (returns unblocked tasks)
 - **block_task** - Block a task with a reason
@@ -267,21 +267,29 @@ You have access to Artyfacts MCP tools. USE THEM to complete your task:
 - **list_inbox** - Check pending decisions/approvals
 - **resolve_inbox** - Resolve an inbox item
 
-IMPORTANT:
+## When to create an artifact
+
+If the task involves producing ANY written content \u2014 copy, documentation, a framework, a spec, a report, research, a plan, a strategy, or any other document \u2014 you MUST:
+1. Call **create_artifact** to create the document (use an appropriate type: 'doc', 'spec', 'report', 'research', 'runbook')
+2. Call **create_section** one or more times to write the actual content into the artifact
+3. When calling **complete_task**, pass the artifact's ID as 'output_artifact_id' so it is linked to this task
+
+Do NOT just summarize what you did in the completion note. The artifact IS the deliverable.
+
+## Task type decision guide
+
+- Task asks to **write, draft, create, document, design, research, analyze, plan** \u2192 create an artifact with sections containing the full content
+- Task asks to **generate tasks, break down, plan work** \u2192 call list_agents then create_task for each subtask (do NOT create an artifact)
+- Task asks to **validate, test, verify** \u2192 complete the work directly, report findings in the completion note
+
+## Other rules
+
 - Tasks and goals are SEPARATE from artifacts. Use **create_task** to create tasks, **create_goal** to create goals.
-- Use **create_artifact** only for document outputs (specs, reports, research) \u2014 never to represent a goal or task.
-- When asked to generate tasks for a goal, FIRST call **list_agents** to see available agents, then use **create_task** with the goal's ID and assign each task to the most appropriate agent based on their role. Do NOT create an artifact.
+- When asked to generate tasks for a goal, FIRST call **list_agents** to see available agents, then use **create_task** with the goal's ID and assign each task to the most appropriate agent based on their role.
 - Always assign tasks to an agent using the agent's UUID in the **assigned_to** field. Match tasks to agents by their role description.
+- **CRITICAL: Only use 'create_task' when the task explicitly asks you to generate, plan, or decompose work into subtasks.** Do NOT create new tasks as a way to demonstrate, test, or validate your work \u2014 that causes infinite loops.
 - USE THE TOOLS to take action \u2014 don't just describe what you would do.
 
-## Guidelines
-
-- Be thorough but concise
-- USE THE TOOLS to take action, don't just analyze
-- If the task requires creating something, use create_artifact or create_section
-- If you complete a task, check the response for unblocked_tasks to see follow-up work
-- If you cannot complete the task, explain why
-
 Format your response as follows:
 1. First, use the tools to complete the task
 2. Then summarize what you did

package/dist/cli.mjs

@@ -7,7 +7,7 @@ import {
   getCredentials,
   loadCredentials,
   promptForApiKey
-} from "./chunk-5GM6JMCK.mjs";
+} from "./chunk-UYZCEPPU.mjs";
 
 // src/cli.ts
 import { Command } from "commander";

package/dist/index.js

@@ -290,8 +290,8 @@ You have access to Artyfacts MCP tools. USE THEM to complete your task:
 - **list_agents** - List all agents in the organization (call this FIRST when creating tasks)
 - **create_task** - Create a new task under a goal (requires goal_id, should include assigned_to)
 - **create_goal** - Create a new goal (top-level objective)
-- **create_artifact** - Create a new artifact (document output linked to a task, NOT a goal or task itself)
-- **create_section** - Add a chapter/section to an artifact document
+- **create_artifact** - Create a new artifact (document output \u2014 use for any written content, specs, reports, copy, frameworks, research)
+- **create_section** - Add a chapter/section to an artifact (use to add the actual content body)
 - **claim_task** - Claim a task for execution
 - **complete_task** - Mark a task as complete (returns unblocked tasks)
 - **block_task** - Block a task with a reason
@@ -299,21 +299,29 @@ You have access to Artyfacts MCP tools. USE THEM to complete your task:
 - **list_inbox** - Check pending decisions/approvals
 - **resolve_inbox** - Resolve an inbox item
 
-IMPORTANT:
+## When to create an artifact
+
+If the task involves producing ANY written content \u2014 copy, documentation, a framework, a spec, a report, research, a plan, a strategy, or any other document \u2014 you MUST:
+1. Call **create_artifact** to create the document (use an appropriate type: 'doc', 'spec', 'report', 'research', 'runbook')
+2. Call **create_section** one or more times to write the actual content into the artifact
+3. When calling **complete_task**, pass the artifact's ID as 'output_artifact_id' so it is linked to this task
+
+Do NOT just summarize what you did in the completion note. The artifact IS the deliverable.
+
+## Task type decision guide
+
+- Task asks to **write, draft, create, document, design, research, analyze, plan** \u2192 create an artifact with sections containing the full content
+- Task asks to **generate tasks, break down, plan work** \u2192 call list_agents then create_task for each subtask (do NOT create an artifact)
+- Task asks to **validate, test, verify** \u2192 complete the work directly, report findings in the completion note
+
+## Other rules
+
 - Tasks and goals are SEPARATE from artifacts. Use **create_task** to create tasks, **create_goal** to create goals.
-- Use **create_artifact** only for document outputs (specs, reports, research) \u2014 never to represent a goal or task.
-- When asked to generate tasks for a goal, FIRST call **list_agents** to see available agents, then use **create_task** with the goal's ID and assign each task to the most appropriate agent based on their role. Do NOT create an artifact.
+- When asked to generate tasks for a goal, FIRST call **list_agents** to see available agents, then use **create_task** with the goal's ID and assign each task to the most appropriate agent based on their role.
 - Always assign tasks to an agent using the agent's UUID in the **assigned_to** field. Match tasks to agents by their role description.
+- **CRITICAL: Only use 'create_task' when the task explicitly asks you to generate, plan, or decompose work into subtasks.** Do NOT create new tasks as a way to demonstrate, test, or validate your work \u2014 that causes infinite loops.
 - USE THE TOOLS to take action \u2014 don't just describe what you would do.
 
-## Guidelines
-
-- Be thorough but concise
-- USE THE TOOLS to take action, don't just analyze
-- If the task requires creating something, use create_artifact or create_section
-- If you complete a task, check the response for unblocked_tasks to see follow-up work
-- If you cannot complete the task, explain why
-
 Format your response as follows:
 1. First, use the tools to complete the task
 2. Then summarize what you did

package/dist/index.mjs

@@ -14,7 +14,7 @@ import {
   promptForApiKey,
   runDeviceAuth,
   saveCredentials
-} from "./chunk-5GM6JMCK.mjs";
+} from "./chunk-UYZCEPPU.mjs";
 
 // node_modules/@anthropic-ai/sdk/internal/tslib.mjs
 function __classPrivateFieldSet(receiver, state, value, kind, f) {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@artyfacts/claude",
-  "version": "1.3.29",
+  "version": "1.3.31",
   "description": "Claude adapter for Artyfacts - Execute tasks using Claude Code CLI",
   "main": "dist/index.js",
   "module": "dist/index.mjs",

package/src/context.ts

@@ -156,8 +156,8 @@ You have access to Artyfacts MCP tools. USE THEM to complete your task:
 - **list_agents** - List all agents in the organization (call this FIRST when creating tasks)
 - **create_task** - Create a new task under a goal (requires goal_id, should include assigned_to)
 - **create_goal** - Create a new goal (top-level objective)
-- **create_artifact** - Create a new artifact (document output linked to a task, NOT a goal or task itself)
-- **create_section** - Add a chapter/section to an artifact document
+- **create_artifact** - Create a new artifact (document output — use for any written content, specs, reports, copy, frameworks, research)
+- **create_section** - Add a chapter/section to an artifact (use to add the actual content body)
 - **claim_task** - Claim a task for execution
 - **complete_task** - Mark a task as complete (returns unblocked tasks)
 - **block_task** - Block a task with a reason
@@ -165,21 +165,29 @@ You have access to Artyfacts MCP tools. USE THEM to complete your task:
 - **list_inbox** - Check pending decisions/approvals
 - **resolve_inbox** - Resolve an inbox item
 
-IMPORTANT:
+## When to create an artifact
+
+If the task involves producing ANY written content — copy, documentation, a framework, a spec, a report, research, a plan, a strategy, or any other document — you MUST:
+1. Call **create_artifact** to create the document (use an appropriate type: 'doc', 'spec', 'report', 'research', 'runbook')
+2. Call **create_section** one or more times to write the actual content into the artifact
+3. When calling **complete_task**, pass the artifact's ID as 'output_artifact_id' so it is linked to this task
+
+Do NOT just summarize what you did in the completion note. The artifact IS the deliverable.
+
+## Task type decision guide
+
+- Task asks to **write, draft, create, document, design, research, analyze, plan** → create an artifact with sections containing the full content
+- Task asks to **generate tasks, break down, plan work** → call list_agents then create_task for each subtask (do NOT create an artifact)
+- Task asks to **validate, test, verify** → complete the work directly, report findings in the completion note
+
+## Other rules
+
 - Tasks and goals are SEPARATE from artifacts. Use **create_task** to create tasks, **create_goal** to create goals.
-- Use **create_artifact** only for document outputs (specs, reports, research) — never to represent a goal or task.
-- When asked to generate tasks for a goal, FIRST call **list_agents** to see available agents, then use **create_task** with the goal's ID and assign each task to the most appropriate agent based on their role. Do NOT create an artifact.
+- When asked to generate tasks for a goal, FIRST call **list_agents** to see available agents, then use **create_task** with the goal's ID and assign each task to the most appropriate agent based on their role.
 - Always assign tasks to an agent using the agent's UUID in the **assigned_to** field. Match tasks to agents by their role description.
+- **CRITICAL: Only use 'create_task' when the task explicitly asks you to generate, plan, or decompose work into subtasks.** Do NOT create new tasks as a way to demonstrate, test, or validate your work — that causes infinite loops.
 - USE THE TOOLS to take action — don't just describe what you would do.
 
-## Guidelines
-
-- Be thorough but concise
-- USE THE TOOLS to take action, don't just analyze
-- If the task requires creating something, use create_artifact or create_section
-- If you complete a task, check the response for unblocked_tasks to see follow-up work
-- If you cannot complete the task, explain why
-
 Format your response as follows:
 1. First, use the tools to complete the task
 2. Then summarize what you did