legion-cc 0.18.0 → 0.19.1

package/agents/legion-orchestrator.md

@@ -170,12 +170,12 @@ Cap at 10.
    Spawn Planner:
    ```
    Agent(team_name: ..., name: "planner", subagent_type: "planner",
-        prompt: "## Assignment\n**Team**: <team-name>\n**Your task ID**: <task-4-id>\n\n## Task\n<the user's original request>\n\n## Phase 1 Bundle Section Task IDs\nRead each Bundle Section using TaskGet(taskId) — the section is in the task's `description` field. Each section starts with `## Bundle Section N:`.\n- **Architect (Bundle Section 1)**: TaskGet('<task-1-id>')\n- **Code Analyst (Bundle Section 2)**: TaskGet('<task-2-id>')\n- **Context Curator (Bundle Section 3)**: TaskGet('<task-3-id>')\n\nRead all three Bundle Sections, assemble the Bundle, then produce an Execution Pack with sub-tasks, pseudo-diffs, acceptance criteria, risks, and rollback. Use ToolSearch('+legion-sequential-thinking') before plan construction. Read Infrastructure Notes sections (§1.6, §2.7) from Bundle Sections. If legion-memory MCP is available, query it for cross-session context via mcp__legion-memory__read_graph. Include User Questions section (≤3) only if critical clarifications are needed.\n\n## End of Work Protocol (MANDATORY)\nTaskUpdate(taskId: '<task-4-id>', status: 'completed', description: '<your full Execution Pack>')\nIf you have user questions, also: SendMessage(type: 'message', recipient: 'legion-orchestrator', content: '<questions>', summary: 'Clarifying questions for user')\nDo NOT go idle without calling TaskUpdate.")
+        prompt: "## Assignment\n**Team**: <team-name>\n**Your task ID**: <task-4-id>\n\n## Task\n<the user's original request>\n\n## Phase 1 Bundle Section Task IDs\nRead each Bundle Section using TaskGet(taskId) — the section is in the task's `description` field. Each section starts with `## Bundle Section N:`.\n- **Architect (Bundle Section 1)**: TaskGet('<task-1-id>')\n- **Code Analyst (Bundle Section 2)**: TaskGet('<task-2-id>')\n- **Context Curator (Bundle Section 3)**: TaskGet('<task-3-id>')\n\nRead all three Bundle Sections, assemble the Bundle, then produce an Execution Pack with sub-tasks, pseudo-diffs, acceptance criteria, risks, and rollback. Use ToolSearch('+legion-sequential-thinking') and ToolSearch('+legion-memory') before plan construction. Read Infrastructure Notes sections (§1.6, §2.7) from Bundle Sections. Query mcp__legion-memory__read_graph for cross-session context. If legion-memory MCP is unavailable, proceed without. Include User Questions section (≤3) only if critical clarifications are needed.\n\n## End of Work Protocol (MANDATORY)\nTaskUpdate(taskId: '<task-4-id>', status: 'completed', description: '<your full Execution Pack>')\nIf you have user questions, also: SendMessage(type: 'message', recipient: 'legion-orchestrator', content: '<questions>', summary: 'Clarifying questions for user')\nDo NOT go idle without calling TaskUpdate.")
    ```
    Spawn Memory (fire-and-forget):
    ```
-   Agent(team_name: ..., name: "memory", subagent_type: "memory",
-        prompt: "## Assignment\n**Team**: <team-name>\n**Your task ID**: <task-6-id>\n\n## Phase 1 Bundle Section Task IDs\nRead Bundle Sections 1 and 2 using TaskGet(taskId) — the section is in the task's `description` field.\n- **Architect (Bundle Section 1)**: TaskGet('<task-1-id>')\n- **Code Analyst (Bundle Section 2)**: TaskGet('<task-2-id>')\n\nExtract key findings and store them in project memory via MCP legion-memory server. Pay special attention to Infrastructure Notes (§1.6, §2.7) — store these as separate named entities using mcp__legion-memory__create_entities and mcp__legion-memory__add_observations.\n\n## End of Work Protocol (MANDATORY)\nTaskUpdate(taskId: '<task-6-id>', status: 'completed', description: '<summary of stored memories>')\nDo NOT go idle without calling TaskUpdate.")
+   Agent(team_name: ..., name: "memory", subagent_type: "general-purpose",
+        prompt: "## Assignment\n**Team**: <team-name>\n**Your task ID**: <task-6-id>\n\n## Phase 1 Bundle Section Task IDs\nRead Bundle Sections 1 and 2 using TaskGet(taskId) — the section is in the task's `description` field.\n- **Architect (Bundle Section 1)**: TaskGet('<task-1-id>')\n- **Code Analyst (Bundle Section 2)**: TaskGet('<task-2-id>')\n\nLoad MCP tools first: ToolSearch(\"+legion-memory\"). Then extract key findings and store them in project memory via MCP legion-memory server. Pay special attention to Infrastructure Notes (§1.6, §2.7) — store these as separate named entities using mcp__legion-memory__create_entities and mcp__legion-memory__add_observations. If MCP is unavailable, note the error and complete anyway.\n\n## End of Work Protocol (MANDATORY)\nTaskUpdate(taskId: '<task-6-id>', status: 'completed', description: '<summary of stored memories>')\nDo NOT go idle without calling TaskUpdate.")
    ```
 6. Wait for plan — poll `TaskList` until tasks 4 and 6 are `status: "completed"`
 
@@ -386,7 +386,7 @@ while not all [t1, t2, t3] completed: check TaskList()
 // Spawn Planner AND Memory in parallel (Phase 2)
 Agent(..., name: "planner", prompt: "## Assignment\n**Your task ID**: " + t4.id + "\n\n## Task\n<the user's original request>\n\n## Phase 1 Bundle Section Task IDs\nRead each Bundle Section using TaskGet(taskId). Each starts with ## Bundle Section N:.\n- Architect (§1): TaskGet('" + t1.id + "')\n- Code Analyst (§2): TaskGet('" + t2.id + "')\n- Context Curator (§3): TaskGet('" + t3.id + "')\n\nAssemble Bundle, produce Execution Pack with optional user questions (≤3).\n\n## End of Work Protocol\nTaskUpdate(taskId: '" + t4.id + "', ...)")
 
-Agent(..., name: "memory", subagent_type: "memory",
+Agent(..., name: "memory", subagent_type: "general-purpose",
   prompt: "## Assignment\n**Your task ID**: " + t6.id + "\n\n## Phase 1 Bundle Section Task IDs\n- Architect (§1): TaskGet('" + t1.id + "')\n- Code Analyst (§2): TaskGet('" + t2.id + "')\n\nExtract key findings and store in project memory via MCP legion-memory server.\n\n## End of Work Protocol\nTaskUpdate(taskId: '" + t6.id + "', ...)")
 
 // Phase 2.5: Check for Planner questions → AskUserQuestion if present

package/agents/memory.md

@@ -49,6 +49,27 @@ You do NOT:
 
 If the legion-memory MCP server is unavailable, note the error in your task description and complete anyway — the pipeline must not be blocked.
 
+## Codebase Mode
+
+When spawned by the **codebase pipeline** (`/legion:codebase`), you read analysis files instead of Bundle Sections:
+
+1. The spawn prompt specifies file paths in `.legion/codebase/` and `.legion/codebase/agents/`
+2. Read each file using the Read tool (not TaskGet)
+3. For each file, create a named entity (`codebase:{name}`) with type `codebase-analysis`
+4. Add condensed observations — key facts, not full file dumps
+5. Create relations between related entities
+6. Use consistent entity names so re-runs are idempotent
+
+Entity naming convention:
+- `codebase:architecture` — from `architecture.md`
+- `codebase:conventions` — from `conventions.md`
+- `codebase:files` — from `files.md`
+- `codebase:dependencies` — from `dependencies.md`
+- `codebase:api` — from `api.md`
+- `codebase:architect-brief` — from `agents/architect-brief.md`
+- `codebase:code-analyst-brief` — from `agents/code-analyst-brief.md`
+- `codebase:context-curator-brief` — from `agents/context-curator-brief.md`
+
 ## MCP Tools
 
 If MCP servers are available in the project, you MUST:

package/bin/install.js

@@ -52,6 +52,8 @@ const libTarget    = path.join(legionHome, 'bin', 'lib');
 const pkgTarget    = path.join(legionHome, 'package.json');
 const settingsPath = path.join(configDir, 'settings.json');
 const claudeJsonPath = path.join(os.homedir(), '.claude.json');
+const memoryDataDir  = path.join(os.homedir(), '.legion');
+const memoryFilePath = path.join(memoryDataDir, 'memory.jsonl');
 const agentsSource   = path.join(projectRoot, 'agents');
 const agentsTarget   = path.join(configDir, 'agents');
 const commandsSource = path.join(projectRoot, 'commands', 'legion');
@@ -714,6 +716,7 @@ What gets installed:
   ~/.claude/legion/package.json              Version metadata
   ~/.claude/settings.json                    Hook registrations (merged)
   ~/.claude.json                             MCP server registrations (merged)
+  ~/.legion/memory.jsonl                     Memory data file (preserved on uninstall)
 `);
 }
 
@@ -803,7 +806,25 @@ function install() {
   log('Registering MCP permissions...');
   registerMcpPermissions();
 
-  // 10. Register MCP servers in ~/.claude.json
+  // 10. Ensure memory data file exists
+  log('Ensuring memory data file...');
+  ensureDir(memoryDataDir);
+  if (flags.dryRun) {
+    log(`${prefix}create ${memoryFilePath} (if not exists)`);
+  } else {
+    try {
+      fs.writeFileSync(memoryFilePath, '', { flag: 'wx', mode: 0o600 });
+      log(`created ${path.basename(memoryFilePath)}`);
+    } catch (err) {
+      if (err.code === 'EEXIST') {
+        log(`skip (exists) ${path.basename(memoryFilePath)}`);
+      } else {
+        throw err;
+      }
+    }
+  }
+
+  // 11. Register MCP servers in ~/.claude.json
   log('Registering MCP servers...');
   registerMcpServers();
 
@@ -870,6 +891,9 @@ function uninstall() {
     }
   } catch (_e) { /* ignore */ }
 
+  // 9. Note: ~/.legion/ is preserved (user data)
+  log('Note: ~/.legion/ preserved (contains user memory data)');
+
   console.log('─'.repeat(40));
   if (flags.dryRun) {
     console.log('Dry-run complete. No files were modified.\n');

package/commands/legion/auto.md

@@ -125,12 +125,12 @@ Escalation: shutdown quick team → create cycle team → run full Cycle Pipelin
    Spawn Planner:
    ```
    Agent(team_name: ..., name: "planner", subagent_type: "planner",
-        prompt: "## Assignment\n**Team**: <team-name>\n**Your task ID**: <task-4-id>\n\n## Task\n<the user's original request>\n\n## Phase 1 Bundle Section Task IDs\nRead each Bundle Section using TaskGet(taskId) — the section is in the task's `description` field. Each section starts with `## Bundle Section N:`.\n- **Architect (Bundle Section 1)**: TaskGet('<task-1-id>')\n- **Code Analyst (Bundle Section 2)**: TaskGet('<task-2-id>')\n- **Context Curator (Bundle Section 3)**: TaskGet('<task-3-id>')\n\nRead all three Bundle Sections, assemble the Bundle, then produce an Execution Pack with sub-tasks, pseudo-diffs, acceptance criteria, risks, and rollback. Use ToolSearch('+legion-sequential-thinking') before plan construction. Read Infrastructure Notes sections (§1.6, §2.7) from Bundle Sections. If legion-memory MCP is available, query it for cross-session context via mcp__legion-memory__read_graph. Include User Questions section (≤3) only if critical clarifications are needed.\n\n## End of Work Protocol (MANDATORY)\nTaskUpdate(taskId: '<task-4-id>', status: 'completed', description: '<your full Execution Pack>')\nIf you have user questions, also: SendMessage(type: 'message', recipient: 'legion-orchestrator', content: '<questions>', summary: 'Clarifying questions for user')\nDo NOT go idle without calling TaskUpdate.")
+        prompt: "## Assignment\n**Team**: <team-name>\n**Your task ID**: <task-4-id>\n\n## Task\n<the user's original request>\n\n## Phase 1 Bundle Section Task IDs\nRead each Bundle Section using TaskGet(taskId) — the section is in the task's `description` field. Each section starts with `## Bundle Section N:`.\n- **Architect (Bundle Section 1)**: TaskGet('<task-1-id>')\n- **Code Analyst (Bundle Section 2)**: TaskGet('<task-2-id>')\n- **Context Curator (Bundle Section 3)**: TaskGet('<task-3-id>')\n\nRead all three Bundle Sections, assemble the Bundle, then produce an Execution Pack with sub-tasks, pseudo-diffs, acceptance criteria, risks, and rollback. Use ToolSearch('+legion-sequential-thinking') and ToolSearch('+legion-memory') before plan construction. Read Infrastructure Notes sections (§1.6, §2.7) from Bundle Sections. Query mcp__legion-memory__read_graph for cross-session context. If legion-memory MCP is unavailable, proceed without. Include User Questions section (≤3) only if critical clarifications are needed.\n\n## End of Work Protocol (MANDATORY)\nTaskUpdate(taskId: '<task-4-id>', status: 'completed', description: '<your full Execution Pack>')\nIf you have user questions, also: SendMessage(type: 'message', recipient: 'legion-orchestrator', content: '<questions>', summary: 'Clarifying questions for user')\nDo NOT go idle without calling TaskUpdate.")
    ```
    Spawn Memory (fire-and-forget):
    ```
-   Agent(team_name: ..., name: "memory", subagent_type: "memory",
-        prompt: "## Assignment\n**Team**: <team-name>\n**Your task ID**: <task-6-id>\n\n## Phase 1 Bundle Section Task IDs\nRead Bundle Sections 1 and 2 using TaskGet(taskId) — the section is in the task's `description` field.\n- **Architect (Bundle Section 1)**: TaskGet('<task-1-id>')\n- **Code Analyst (Bundle Section 2)**: TaskGet('<task-2-id>')\n\nExtract key findings and store them in project memory via MCP legion-memory server. Pay special attention to Infrastructure Notes (§1.6, §2.7) — store these as separate named entities using mcp__legion-memory__create_entities and mcp__legion-memory__add_observations.\n\n## End of Work Protocol (MANDATORY)\nTaskUpdate(taskId: '<task-6-id>', status: 'completed', description: '<summary of stored memories>')\nDo NOT go idle without calling TaskUpdate.")
+   Agent(team_name: ..., name: "memory", subagent_type: "general-purpose",
+        prompt: "## Assignment\n**Team**: <team-name>\n**Your task ID**: <task-6-id>\n\n## Phase 1 Bundle Section Task IDs\nRead Bundle Sections 1 and 2 using TaskGet(taskId) — the section is in the task's `description` field.\n- **Architect (Bundle Section 1)**: TaskGet('<task-1-id>')\n- **Code Analyst (Bundle Section 2)**: TaskGet('<task-2-id>')\n\nLoad MCP tools first: ToolSearch(\"+legion-memory\"). Then extract key findings and store them in project memory via MCP legion-memory server. Pay special attention to Infrastructure Notes (§1.6, §2.7) — store these as separate named entities using mcp__legion-memory__create_entities and mcp__legion-memory__add_observations. If MCP is unavailable, note the error and complete anyway.\n\n## End of Work Protocol (MANDATORY)\nTaskUpdate(taskId: '<task-6-id>', status: 'completed', description: '<summary of stored memories>')\nDo NOT go idle without calling TaskUpdate.")
    ```
 6. Wait for plan — poll `TaskList` until tasks 4 and 6 are `status: "completed"`
 

package/commands/legion/codebase.md

@@ -11,7 +11,7 @@ $ARGUMENTS
 
 1. **Create team**: `TeamCreate(team_name: "legion-codebase-<timestamp>")`
 2. **Create output directories**: Run `mkdir -p "$PWD/.legion/codebase/agents/"` via Bash
-3. **Create 8 tasks** in two phases:
+3. **Create 9 tasks** in three phases:
    **Phase 1** (all parallel, no dependencies):
    - Task 1: "Analyze architecture" → architecture-analyst
    - Task 2: "Analyze conventions" → conventions-analyst
@@ -22,6 +22,8 @@ $ARGUMENTS
    - Task 6: "Write architect brief" → architect-brief-writer (blockedBy: [1,2,3,4,5])
    - Task 7: "Write code-analyst brief" → code-analyst-brief-writer (blockedBy: [1,2,3,4,5])
    - Task 8: "Write context-curator brief" → context-curator-brief-writer (blockedBy: [1,2,3,4,5])
+   **Phase 3** (blockedBy Phase 1+2):
+   - Task 9: "Store codebase analysis in MCP memory" → memory-agent (blockedBy: [1,2,3,4,5,6,7,8])
 4. **Phase 1 — Spawn 5 teammates in parallel**:
    - `architecture-analyst` (subagent_type: "architect") — analyze project architecture: modules, data flow, key design decisions, component relationships. Write results to `.legion/codebase/architecture.md` via Write tool.
    - `conventions-analyst` (subagent_type: "context-curator") — analyze code style, naming conventions, patterns, project rules, formatting. Write results to `.legion/codebase/conventions.md` via Write tool.
@@ -34,7 +36,47 @@ $ARGUMENTS
    - `code-analyst-brief-writer` (subagent_type: "code-analyst") — read `.legion/codebase/files.md`, `.legion/codebase/dependencies.md`, `.legion/codebase/api.md`. Write a concise code-analyst brief (plain markdown, no YAML frontmatter, ~50-100 lines) to `.legion/codebase/agents/code-analyst-brief.md` via Write tool. Structure: `# Code Analyst Brief` → Quick Reference table → File Map (directory tree with purposes) → Entry Points (CLI, hooks, libs) → Import Graph (condensed require() chains) → Key Callsites (settings.json registrations, HOOK_FILES/AGENT_FILES arrays) → Runtime File Paths (/tmp/legion-* with reader/writer) → Test Status (none, manual validation). Dense facts, minimal prose.
    - `context-curator-brief-writer` (subagent_type: "context-curator") — read `.legion/codebase/conventions.md`, `.legion/codebase/architecture.md`, `.legion/codebase/files.md`. Write a concise context-curator brief (plain markdown, no YAML frontmatter, ~50-100 lines) to `.legion/codebase/agents/context-curator-brief.md` via Write tool. Structure: `# Context Curator Brief` → Quick Reference table → Project Identity (name, type, language, no frameworks) → Rules & Conventions (module format, zero deps, atomic writes, silent fail, naming) → Build/Run/Test table (install, force, dry-run, uninstall — no lint/test) → Config Files (config.json, package.json, settings.json) → Environment (CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1, Node >= 18) → Project-specific Patterns → Key Files table. Dense facts, minimal prose.
 7. **Wait** for Phase 2 (all 3) to complete
-8. **Verify** all 8 output files exist:
+8. **Phase 3 — Spawn memory agent**:
+   - `memory-agent` (subagent_type: "general-purpose") — read all generated files from `.legion/codebase/` and `.legion/codebase/agents/`, extract key findings, store in MCP legion-memory as named entities.
+
+   Spawn prompt:
+   ```
+   ## Assignment
+   **Team**: <team-name>
+   **Your task ID**: <Task 9 id>
+   **Mode**: Codebase analysis storage
+
+   ## Source Files
+   Read each file using Read tool:
+   - `.legion/codebase/architecture.md` — project architecture, modules, data flow
+   - `.legion/codebase/conventions.md` — code style, naming, patterns
+   - `.legion/codebase/files.md` — file map with descriptions
+   - `.legion/codebase/dependencies.md` — module dependencies, import graph
+   - `.legion/codebase/api.md` — public interfaces, CLI, hooks protocol
+   - `.legion/codebase/agents/architect-brief.md` — architect quick-reference
+   - `.legion/codebase/agents/code-analyst-brief.md` — code-analyst quick-reference
+   - `.legion/codebase/agents/context-curator-brief.md` — context-curator quick-reference
+
+   ## Instructions
+   1. Load MCP tools: `ToolSearch("+legion-memory")`
+   2. Read all 8 files listed above
+   3. For each file, create a named entity using `mcp__legion-memory__create_entities`:
+      - Entity name: `codebase:{filename}` (e.g., `codebase:architecture`, `codebase:conventions`)
+      - Entity type: `codebase-analysis`
+   4. For each entity, add key observations (condensed summaries, not full file content) using `mcp__legion-memory__add_observations`
+   5. Create relations between entities using `mcp__legion-memory__create_relations`:
+      - `codebase:architecture` → `relates_to` → `codebase:dependencies`
+      - `codebase:api` → `relates_to` → `codebase:files`
+      - `codebase:conventions` → `relates_to` → `codebase:architecture`
+      - Brief entities → `summarizes` → their source entities
+   6. If MCP is unavailable, note the error and complete anyway
+
+   ## End of Work Protocol (MANDATORY)
+   TaskUpdate(taskId: "<task-9-id>", status: "completed", description: "<summary of entities stored>")
+   Do NOT go idle without calling TaskUpdate.
+   ```
+9. **Wait** for Phase 3 (Task 9) to complete
+10. **Verify** all 8 output files exist:
    In `.legion/codebase/`:
    - `architecture.md`
    - `conventions.md`
@@ -45,8 +87,8 @@ $ARGUMENTS
    - `architect-brief.md`
    - `code-analyst-brief.md`
    - `context-curator-brief.md`
-9. **Clean shutdown** — shutdown all teammates, TeamDelete
-10. **Report** — summarize what was generated: 5 codebase analysis files in `.legion/codebase/` and 3 agent briefs in `.legion/codebase/agents/`
+11. **Clean shutdown** — shutdown all teammates, TeamDelete
+12. **Report** — summarize what was generated: 5 codebase analysis files in `.legion/codebase/`, 3 agent briefs in `.legion/codebase/agents/`, and key findings stored in MCP memory
 
 ## Rules
 

package/commands/legion/cycle.md

@@ -46,7 +46,7 @@ $ARGUMENTS
 
 5. **Spawn Planner AND Memory in parallel**:
    - **Planner** (subagent_type: "planner") with Phase 1 Bundle Section task IDs and Task 4's ID:
-   - **Memory** (subagent_type: "memory") with Phase 1 Bundle Section task IDs (1, 2) and Task 6's ID — fire-and-forget, does NOT block Implementer
+   - **Memory** (subagent_type: "general-purpose") with Phase 1 Bundle Section task IDs (1, 2) and Task 6's ID — fire-and-forget, does NOT block Implementer
 
    Spawn Planner:
    ```
@@ -63,7 +63,7 @@ $ARGUMENTS
    - **Code Analyst (Bundle Section 2)**: TaskGet('<task-2-id>')
    - **Context Curator (Bundle Section 3)**: TaskGet('<task-3-id>')
 
-   Read all three Bundle Sections, assemble the Bundle, then produce an Execution Pack with sub-tasks, pseudo-diffs, acceptance criteria, risks, and rollback. Use ToolSearch('+legion-sequential-thinking') before plan construction. Read Infrastructure Notes sections (§1.6, §2.7) from Bundle Sections. If legion-memory MCP is available, query it for cross-session context via mcp__legion-memory__read_graph. Include User Questions section (≤3) only if critical clarifications are needed.
+   Read all three Bundle Sections, assemble the Bundle, then produce an Execution Pack with sub-tasks, pseudo-diffs, acceptance criteria, risks, and rollback. Use ToolSearch('+legion-sequential-thinking') and ToolSearch('+legion-memory') before plan construction. Read Infrastructure Notes sections (§1.6, §2.7) from Bundle Sections. Query mcp__legion-memory__read_graph for cross-session context. If legion-memory MCP is unavailable, proceed without. Include User Questions section (≤3) only if critical clarifications are needed.
 
    ## End of Work Protocol (MANDATORY)
    TaskUpdate(taskId: "<your-task-id>", status: "completed", description: "<your full Execution Pack>")
@@ -72,8 +72,8 @@ $ARGUMENTS
    ```
    Spawn Memory (fire-and-forget):
    ```
-   Agent(team_name: ..., name: "memory", subagent_type: "memory",
-        prompt: "## Assignment\n**Team**: <team-name>\n**Your task ID**: <Task 6 id>\n\n## Phase 1 Bundle Section Task IDs\nRead Bundle Sections 1 and 2 using TaskGet(taskId) — the section is in the task's `description` field.\n- **Architect (Bundle Section 1)**: TaskGet('<task-1-id>')\n- **Code Analyst (Bundle Section 2)**: TaskGet('<task-2-id>')\n\nExtract key findings and store them in project memory via MCP legion-memory server. Pay special attention to Infrastructure Notes (§1.6, §2.7) — store these as separate named entities using mcp__legion-memory__create_entities and mcp__legion-memory__add_observations.\n\n## End of Work Protocol (MANDATORY)\nTaskUpdate(taskId: '<task-6-id>', status: 'completed', description: '<summary of stored memories>')\nDo NOT go idle without calling TaskUpdate.")
+   Agent(team_name: ..., name: "memory", subagent_type: "general-purpose",
+        prompt: "## Assignment\n**Team**: <team-name>\n**Your task ID**: <Task 6 id>\n\n## Phase 1 Bundle Section Task IDs\nRead Bundle Sections 1 and 2 using TaskGet(taskId) — the section is in the task's `description` field.\n- **Architect (Bundle Section 1)**: TaskGet('<task-1-id>')\n- **Code Analyst (Bundle Section 2)**: TaskGet('<task-2-id>')\n\nLoad MCP tools first: ToolSearch(\"+legion-memory\"). Then extract key findings and store them in project memory via MCP legion-memory server. Pay special attention to Infrastructure Notes (§1.6, §2.7) — store these as separate named entities using mcp__legion-memory__create_entities and mcp__legion-memory__add_observations. If MCP is unavailable, note the error and complete anyway.\n\n## End of Work Protocol (MANDATORY)\nTaskUpdate(taskId: '<task-6-id>', status: 'completed', description: '<summary of stored memories>')\nDo NOT go idle without calling TaskUpdate.")
    ```
 6. **Wait** for plan — poll `TaskList` until tasks 4 and 6 are `status: "completed"`
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "legion-cc",
-  "version": "0.18.0",
+  "version": "0.19.1",
   "description": "Legion — context monitoring hooks and agent teams for Claude Code",
   "type": "commonjs",
   "bin": {