@axiom-lattice/core 2.1.83 → 2.1.84

package/dist/index.mjs

@@ -20330,7 +20330,7 @@ You have nine tools for agent management:
 - **list_tools** \u2014 See all available tools that can be assigned to agents
 - **get_agent** \u2014 View the full configuration of a specific agent
 - **create_agent** \u2014 Create a REACT or DEEP_AGENT agent
-- **create_workflow** \u2014 Create a WORKFLOW agent from a concise DSL (load create-workflow skill first)
+- **create_workflow** \u2014 Create a WORKFLOW agent from a concise DSL (load create-workflow skill + task-relevant domain skills first)
 - **validate_workflow** \u2014 Validate a workflow agent's DSL
 - **update_workflow** \u2014 Update a workflow agent's DSL or config
 - **update_agent** \u2014 Modify an existing REACT or DEEP_AGENT agent's configuration
@@ -20427,14 +20427,23 @@ Use this when the process is fully known. A workflow is a deterministic LangGrap
 | needs + if model (YAML DSL) | Single orchestrator delegates linearly |
 | No LLM routing decisions | Orchestrator uses LLM to route |
 
-### Phase 0: Load the Skill
+### Phase 0: Load Skills
 
-**BEFORE designing, you MUST load the create-workflow skill:**
+**BEFORE designing, you MUST load relevant skills:**
+
+1. **Always load** the \`create-workflow\` skill \u2014 it teaches the YAML DSL syntax, step format, and design patterns:
 
 \`\`\`
 skill(skill_name: "create-workflow")
 \`\`\`
 
+2. **Determine and load task-relevant domain skills** \u2014 do NOT skip this step:
+   a) **First, scan** the \`<available_skills>\` section above. For each skill, read its name and description to determine if it relates to the user's task domain (e.g., financial analysis, sales, data processing, reporting)
+   b) **Then, load each relevant skill** by calling \`skill(skill_name: "<skill-name>")\` \u2014 call this once per relevant skill, NOT in a batch
+   c) Read the loaded skill content for domain-specific workflow patterns, reusable sub-workflows, and DSL best practices
+
+**If you skip step 2, you will miss critical domain knowledge and produce a suboptimal design.**
+
 ### Phase 1: Design
 
 1. **Analyze the process.** Map every step, branch, data dependency.
@@ -20549,7 +20558,7 @@ All fields except name, type, and prompt are optional.
 
 ### create_workflow (WORKFLOW)
 
-Creates a WORKFLOW agent from a YAML linear DSL. Before calling, load the \`create-workflow\` skill. Must pass \`skillLoaded: true\`.
+Creates a WORKFLOW agent from a YAML linear DSL. Before calling, load the \`create-workflow\` skill plus any task-relevant domain skills. Must pass \`skillLoaded: true\`.
 
 \`\`\`typescript
 {
@@ -20864,6 +20873,14 @@ var agentArchitectConfig = {
   ],
   internalSubAgents: [agentReviewerConfig],
   middleware: [
+    {
+      id: "filesystem",
+      type: "filesystem",
+      name: "Filesystem",
+      description: "Read, write, edit, list, search files",
+      enabled: true,
+      config: {}
+    },
     {
       id: "skill",
       type: "skill",