@newsails/veil-cli 1.0.1

package/.veil/agents/analyst/AGENT.md

@@ -0,0 +1,21 @@
+You are Analyst, a thorough file and project analysis agent.
+
+Your mission: explore project directories, read files, and produce structured analysis reports.
+
+Capabilities:
+- list_dir: explore directory structures
+- read_file: read any file content
+- glob: find files by pattern
+- grep: search file contents
+- memory_write: save important findings for later
+- todo_write/todo_read: manage analysis checklist
+
+Approach every analysis task systematically:
+1. Use todo_write to create an analysis checklist
+2. Explore the directory structure with list_dir
+3. Read key files (README, config, entry points)
+4. Use grep to find patterns across the codebase
+5. Write findings to memory as you go
+6. Produce a final structured report
+
+Be thorough. Cover: project purpose, structure, technologies, key components, and observations.

package/.veil/agents/analyst/agent.json

@@ -0,0 +1,23 @@
+{
+  "name": "analyst",
+  "description": "File and directory analysis agent. Explores project structures, reads files, summarizes codebases.",
+  "model": "moonshotai/kimi-k2.5",
+  "temperature": 0.7,
+  "reasoning": "medium",
+  "memory": { "enabled": false },
+  "skillDiscovery": false,
+  "modes": {
+    "chat": {
+      "enabled": true,
+      "tools": ["read_file", "list_dir", "glob", "grep", "memory_write", "memory_read", "todo_write", "todo_read"],
+      "disallowedTools": ["bash", "write_file", "edit_file"]
+    },
+    "task": {
+      "enabled": true,
+      "maxIterations": 30,
+      "maxDurationSeconds": 180,
+      "tools": ["read_file", "list_dir", "glob", "grep", "memory_write", "memory_read", "todo_write", "todo_read"],
+      "disallowedTools": ["bash", "write_file", "edit_file"]
+    }
+  }
+}

package/.veil/agents/assistant/AGENT.md

@@ -0,0 +1,15 @@
+You are Assistant, a capable general-purpose AI agent.
+
+You can help with a wide range of tasks including:
+- Answering questions and explaining concepts
+- Reading and writing files
+- Analyzing codebases
+- Running shell commands (when safe and necessary)
+- Summarizing documents
+
+When given a task:
+1. Think through what you need to do
+2. Use the appropriate tools (read_file, write_file, bash, etc.)
+3. Report your findings or completion clearly
+
+Be concise in your responses. Prefer action over lengthy explanation.

package/.veil/agents/assistant/agent.json

@@ -0,0 +1,19 @@
+{
+  "name": "assistant",
+  "description": "General-purpose assistant with file and shell access. Handles both chat and tasks.",
+  "model": "moonshotai/kimi-k2.5",
+  "temperature": 0.7,
+  "reasoning": "medium",
+  "memory": { "enabled": false },
+  "skillDiscovery": false,
+  "modes": {
+    "chat": { "enabled": true },
+    "task": {
+      "enabled": true,
+      "maxIterations": 20,
+      "maxDurationSeconds": 120,
+      "tools": ["bash", "read_file", "write_file", "edit_file", "list_dir", "glob", "grep"],
+      "disallowedTools": []
+    }
+  }
+}

package/.veil/agents/coder/AGENT.md

@@ -0,0 +1,18 @@
+You are Coder, an expert code analysis agent.
+
+Your specialties:
+- Reading and understanding codebases
+- Explaining what code does and why
+- Identifying bugs, anti-patterns, and security issues
+- Suggesting improvements with clear rationale
+- Documenting functions and modules
+
+You have READ-ONLY access (no writes, no bash). Use read_file, list_dir, glob, and grep to explore codebases.
+
+When analyzing code:
+1. Start by listing the directory structure
+2. Read key files (entry points, config, core modules)
+3. Identify patterns and architecture
+4. Provide clear, structured analysis
+
+Be precise and technical. Reference specific line numbers and file paths.

package/.veil/agents/coder/agent.json

@@ -0,0 +1,19 @@
+{
+  "name": "coder",
+  "description": "Code analysis and explanation agent. Reads codebases, explains code, finds bugs, suggests improvements.",
+  "model": "moonshotai/kimi-k2.5",
+  "temperature": 0.7,
+  "reasoning": "medium",
+  "memory": { "enabled": false },
+  "skillDiscovery": false,
+  "modes": {
+    "chat": { "enabled": true },
+    "task": {
+      "enabled": true,
+      "maxIterations": 25,
+      "maxDurationSeconds": 180,
+      "tools": ["read_file", "list_dir", "glob", "grep"],
+      "disallowedTools": ["bash", "write_file", "edit_file"]
+    }
+  }
+}

package/.veil/agents/hello/AGENT.md

@@ -0,0 +1,5 @@
+You are Hello, a friendly conversational agent.
+
+Keep responses concise and warm. You can discuss any topic, answer questions, and engage in casual conversation. You have no access to tools — just respond from your knowledge.
+
+When greeted, greet back warmly. When asked a question, answer directly and helpfully.

package/.veil/agents/hello/agent.json

@@ -0,0 +1,13 @@
+{
+  "name": "hello",
+  "description": "Simple conversational agent. No tools. Answers questions, chats, and greets.",
+  "model": "moonshotai/kimi-k2.5",
+  "temperature": 0.7,
+  "reasoning": "medium",
+  "memory": { "enabled": false },
+  "skillDiscovery": false,
+  "modes": {
+    "chat": { "enabled": true },
+    "task": { "enabled": true, "maxIterations": 5, "maxDurationSeconds": 60 }
+  }
+}

package/.veil/agents/writer/AGENT.md

@@ -0,0 +1,12 @@
+You are Writer, a creative writing and content generation agent.
+
+You excel at:
+- Writing engaging stories, articles, and blog posts
+- Summarizing and rewriting content
+- Creating structured documents (reports, READMEs, specs)
+- Adapting tone and style to the audience
+- Generating creative content on any topic
+
+You work purely from knowledge — no file access needed. Just produce high-quality written content.
+
+Always deliver complete, polished output. Don't just outline — write the full content requested.

package/.veil/agents/writer/agent.json

@@ -0,0 +1,17 @@
+{
+  "name": "writer",
+  "description": "Creative writing agent. Generates stories, articles, summaries, and structured documents. No tools needed.",
+  "model": "moonshotai/kimi-k2.5",
+  "temperature": 0.9,
+  "reasoning": "medium",
+  "memory": { "enabled": false },
+  "skillDiscovery": false,
+  "modes": {
+    "chat": { "enabled": true },
+    "task": {
+      "enabled": true,
+      "maxIterations": 8,
+      "maxDurationSeconds": 90
+    }
+  }
+}

package/.veil/memory/MEMORY.md

@@ -0,0 +1,343 @@
+
+
+<!-- 2026-03-02 -->
+# Project Analysis: VeilCli
+
+## Basic Information
+- **Name**: veilcli
+- **Version**: 0.1.0
+- **Description**: Enterprise autonomous agent runtime — local, API-first, multi-model
+- **License**: MIT
+- **Node.js Requirements**: >=18.3.0
+
+## Entry Points
+- **CLI Binary**: `./cli/index.js` (accessible via `veil` command)
+- **Start Command**: `node cli/index.js start`
+- **Test Command**: `node test/smoke.js`
+
+## Key Dependencies
+- **ajv** (8.17.1): JSON schema validation
+- **ajv-formats** (3.0.1): Additional format validators for AJV
+- **better-sqlite3** (11.10.0): SQLite database (synchronous, high-performance)
+- **cors** (2.8.5): Cross-Origin Resource Sharing middleware
+- **express** (4.21.2): Web framework for API
+- **node-cron** (3.0.3): Cron job scheduling
+
+## Architecture Insights
+- API-first design with Express.js
+- Local SQLite database for persistence
+- Supports scheduled tasks via node-cron
+- JSON schema validation for structured data
+- Multi-model agent runtime capability
+
+## Notes
+- No README.md file exists in the project root
+- No devDependencies defined
+- Project appears to be in early development (v0.1.0)
+
+<!-- 2026-03-02 -->
+<!-- 2026-03-02 -->
+## Complete Project Analysis: VeilCLI
+
+### Project Overview
+**VeilCLI** is an enterprise-grade autonomous agent runtime built in Node.js. It provides a local-first, API-driven platform for running AI agents with multi-model support.
+
+### JavaScript File Count
+Approximately **68 JavaScript files** in the main codebase (excluding node_modules and .OLD directory):
+- tools/: 25 files
+- TEST-APP/: 15 files  
+- core/: 8 files
+- api/: 8 files
+- cli/: 3 files
+- utils/: 4 files
+- infrastructure/: 2 files
+- llm/: 1 file
+- settings/: 1 file
+- test/: 1 file
+
+### Core Architecture Components
+
+#### 1. Agent System (core/agent.js)
+- JSON-schema validated agent definitions
+- Hierarchical agent loading (project-level → global)
+- Multi-mode support: chat, task, daemon, subagent
+- Agent configuration via agent.json + AGENT.md + SOUL.md
+
+#### 2. Execution Loop (core/loop.js)
+- Async generator-based agentic loop
+- Tool permission system (allow/ask/deny)
+- Anti-drift reminders every 10 iterations
+- Stall detection and recovery
+- Token tracking and limit enforcement
+
+#### 3. Tool Registry (core/registry.js)
+- 25 built-in tools organized by category:
+  - Agent communication (agent_message, agent_send, agent_spawn)
+  - File operations (read_file, write_file, edit_file, glob, list_dir)
+  - Code search (grep)
+  - System (bash, sleep)
+  - Task management (5 tools)
+  - Memory (3 tools)
+  - Planning (todo_read, todo_write)
+  - Web (web_search, web_fetch)
+  - Utilities (tool_search, log_write)
+- Custom tool support via tool.json + index.js
+- Input validation via AJV
+
+#### 4. Database Layer (infrastructure/database.js)
+- SQLite with better-sqlite3
+- WAL mode enabled
+- Migration system
+- Stores: sessions, tasks, messages, todos, agent-messages
+
+#### 5. LLM Client (llm/client.js)
+- Native fetch-based (no SDK dependency)
+- OpenAI-compatible API support
+- Token usage tracking
+
+#### 6. API Server (api/server.js)
+- Express.js with CORS
+- Route structure: /agents, /sessions, /tasks, /daemons, /system
+- Auth middleware
+
+#### 7. CLI Interface (cli/index.js)
+- Commands: start, stop, run, status, agents, login
+- Auto-daemon startup
+- Task execution mode
+
+### Key Dependencies
+- **express** (4.21.2): Web framework
+- **better-sqlite3** (11.10.0): High-performance SQLite
+- **ajv** (8.17.1): JSON schema validation
+- **node-cron** (3.0.3): Cron scheduling
+- **cors** (2.8.5): Cross-origin middleware
+
+### Notable Patterns
+
+1. **Mode-Based Execution**: Agents run in chat, task, daemon, or subagent modes with different behaviors
+2. **Permission Hierarchy**: Tool permissions checked at agent-level, then settings-level (deny → ask → allow)
+3. **Safety-First**: Auth file access is hardcoded as forbidden; permission system prevents unauthorized operations
+4. **Hot Reload**: Custom tools support cache-busting for development
+5. **Graceful Degradation**: LLM errors are retried up to 3 times with backoff
+6. **Context Compaction**: Anti-drift reminders and context management to keep agents focused
+7. **Multi-Level Configuration**: Project-level settings override global settings
+8. **Async Generators**: Core loop yields progress events for real-time monitoring
+
+### Directory Structure
+```
+VeilCLI/
+├── api/           # Express routes and middleware
+├── cli/           # Command-line interface
+├── core/          # Agent runtime, loop, registry, router
+├── infrastructure/# Database and scheduler
+├── llm/           # LLM client abstraction
+├── settings/      # Configuration field definitions
+├── tools/         # Built-in tool implementations
+├── utils/         # Path resolution, context, settings loader
+├── schemas/       # JSON schemas for validation
+├── migrations/    # Database migrations
+├── system-prompts/# Layered prompt system
+├── examples/      # Example agent configurations
+└── test/          # Smoke tests
+```
+
+### Configuration Locations
+- **Project-level**: `.veil/` folder in project root
+- **Global**: `~/.veil/` in user's home directory
+- **Priority**: Project settings override global settings
+
+### Security Features
+- Auth file reading is explicitly blocked
+- Permission patterns support wildcards
+- Tool input validation via JSON schema
+- PreToolUse and PostToolUse hooks
+- Permission levels: deny, ask, allow
+
+<!-- 2026-03-02 -->
+<!-- 2026-03-02 -->
+## Project Metadata: VeilCLI
+
+### Basic Information
+- **Name**: veilcli
+- **Version**: 0.1.0
+- **Description**: Enterprise autonomous agent runtime — local, API-first, multi-model
+- **License**: MIT
+- **Node.js Requirements**: >=18.3.0
+
+### Entry Points
+- **CLI Binary**: `./cli/index.js` (accessible via `veil` command)
+- **Start Command**: `node cli/index.js start`
+- **Test Command**: `node test/smoke.js`
+
+### Key Dependencies
+- **ajv** (8.17.1): JSON schema validation
+- **ajv-formats** (3.0.1): Additional format validators for AJV
+- **better-sqlite3** (11.10.0): SQLite database (synchronous, high-performance)
+- **cors** (2.8.5): Cross-Origin Resource Sharing middleware
+- **express** (4.21.2): Web framework for API
+- **node-cron** (3.0.3): Cron job scheduling
+
+### Architecture Insights
+- API-first design with Express.js
+- Local SQLite database for persistence
+- Supports scheduled tasks via node-cron
+- JSON schema validation for structured data
+- Multi-model agent runtime capability
+
+### Notes
+- No README.md file exists in the project root
+- No devDependencies defined
+- Project appears to be in early development (v0.1.0)
+
+<!-- 2026-03-02 -->
+## Project Metadata: VeilCLI
+
+### Basic Information
+- **Name**: veilcli
+- **Version**: 0.1.0
+- **Description**: Enterprise autonomous agent runtime — local, API-first, multi-model
+- **License**: MIT
+- **Node.js Requirements**: >=18.3.0
+
+### Entry Points
+- **CLI Binary**: `./cli/index.js` (accessible via `veil` command)
+- **Start Command**: `node cli/index.js start`
+- **Test Command**: `node test/smoke.js`
+
+### Key Dependencies
+- **ajv** (8.17.1): JSON schema validation
+- **ajv-formats** (3.0.1): Additional format validators for AJV
+- **better-sqlite3** (11.10.0): SQLite database (synchronous, high-performance)
+- **cors** (2.8.5): Cross-Origin Resource Sharing middleware
+- **express** (4.21.2): Web framework for API
+- **node-cron** (3.0.3): Cron job scheduling
+
+### Architecture Insights
+- API-first design with Express.js
+- Local SQLite database for persistence
+- Supports scheduled tasks via node-cron
+- JSON schema validation for structured data
+- Multi-model agent runtime capability
+
+### Notes
+- No README.md file exists in the project root
+- No devDependencies defined
+- Project appears to be in early development (v0.1.0)
+
+<!-- 2026-03-02 -->
+<!-- 2026-03-02 -->
+## Complete Project Analysis: VeilCLI
+
+### Project Overview
+**VeilCLI** is an enterprise-grade autonomous agent runtime built in Node.js. It provides a local-first, API-driven platform for running AI agents with multi-model support.
+
+### JavaScript File Count
+**78 JavaScript files** in the main codebase (excluding node_modules and .OLD directory):
+- tools/: 25 files
+- TEST-APP/: 16 files  
+- core/: 8 files
+- api/: 7 files
+- cli/: 3 files
+- utils/: 4 files
+- infrastructure/: 2 files
+- llm/: 1 file
+- settings/: 1 file
+- test/: 1 file
+
+### Core Architecture Components
+
+#### 1. Agent System (core/agent.js)
+- JSON-schema validated agent definitions
+- Hierarchical agent loading (project-level → global)
+- Multi-mode support: chat, task, daemon, subagent
+- Agent configuration via agent.json + AGENT.md + SOUL.md
+
+#### 2. Execution Loop (core/loop.js)
+- Async generator-based agentic loop
+- Tool permission system (allow/ask/deny)
+- Anti-drift reminders every 10 iterations
+- Stall detection and recovery
+- Token tracking and limit enforcement
+
+#### 3. Tool Registry (core/registry.js)
+- 25 built-in tools organized by category:
+  - Agent communication (agent_message, agent_send, agent_spawn)
+  - File operations (read_file, write_file, edit_file, glob, list_dir)
+  - Code search (grep)
+  - System (bash, sleep)
+  - Task management (5 tools)
+  - Memory (3 tools)
+  - Planning (todo_read, todo_write)
+  - Web (web_search, web_fetch)
+  - Utilities (tool_search, log_write)
+- Custom tool support via tool.json + index.js
+- Input validation via AJV
+
+#### 4. Database Layer (infrastructure/database.js)
+- SQLite with better-sqlite3
+- WAL mode enabled
+- Migration system
+- Stores: sessions, tasks, messages, todos, agent-messages
+
+#### 5. LLM Client (llm/client.js)
+- Native fetch-based (no SDK dependency)
+- OpenAI-compatible API support
+- Token usage tracking
+
+#### 6. API Server (api/server.js)
+- Express.js with CORS
+- Route structure: /agents, /sessions, /tasks, /daemons, /system
+- Auth middleware
+
+#### 7. CLI Interface (cli/index.js)
+- Commands: start, stop, run, status, agents, login
+- Auto-daemon startup
+- Task execution mode
+
+### Key Dependencies
+- **express** (4.21.2): Web framework
+- **better-sqlite3** (11.10.0): High-performance SQLite
+- **ajv** (8.17.1): JSON schema validation
+- **node-cron** (3.0.3): Cron scheduling
+- **cors** (2.8.5): Cross-origin middleware
+
+### Notable Patterns
+
+1. **Mode-Based Execution**: Agents run in chat, task, daemon, or subagent modes with different behaviors
+2. **Permission Hierarchy**: Tool permissions checked at agent-level, then settings-level (deny → ask → allow)
+3. **Safety-First**: Auth file access is hardcoded as forbidden; permission system prevents unauthorized operations
+4. **Hot Reload**: Custom tools support cache-busting for development
+5. **Graceful Degradation**: LLM errors are retried up to 3 times with backoff
+6. **Context Compaction**: Anti-drift reminders and context management to keep agents focused
+7. **Multi-Level Configuration**: Project-level settings override global settings
+8. **Async Generators**: Core loop yields progress events for real-time monitoring
+
+### Directory Structure
+```
+VeilCLI/
+├── api/           # Express routes and middleware
+├── cli/           # Command-line interface
+├── core/          # Agent runtime, loop, registry, router
+├── infrastructure/# Database and scheduler
+├── llm/           # LLM client abstraction
+├── settings/      # Configuration field definitions
+├── tools/         # Built-in tool implementations
+├── utils/         # Path resolution, context, settings loader
+├── schemas/       # JSON schemas for validation
+├── migrations/    # Database migrations
+├── system-prompts/# Layered prompt system
+├── examples/      # Example agent configurations
+└── test/          # Smoke tests
+```
+
+### Configuration Locations
+- **Project-level**: `.veil/` folder in project root
+- **Global**: `~/.veil/` in user's home directory
+- **Priority**: Project settings override global settings
+
+### Security Features
+- Auth file reading is explicitly blocked
+- Permission patterns support wildcards
+- Tool input validation via JSON schema
+- PreToolUse and PostToolUse hooks
+- Permission levels: deny, ask, allow

package/.veil/memory/agents/analyst/MEMORY.md

@@ -0,0 +1,55 @@
+
+
+<!-- 2026-03-02 -->
+## Tools Directory Analysis
+
+The `/home/ixi/khacloud/drive/Plugins/VeilCli/tools` directory contains **25 tool files** that provide the core functionality for VeilCLI agents.
+
+### Tool Categories:
+
+**Agent Communication (3 tools):**
+- `agent_message.js` - Synchronous message to another agent with response waiting
+- `agent_send.js` - Asynchronous fire-and-forget message to another agent
+- `agent_spawn.js` - Spawn sub-agents with sync/async modes and return modes
+
+**File Operations (5 tools):**
+- `read_file.js` - Read file contents with offset/limit support
+- `write_file.js` - Write/overwrite entire files
+- `edit_file.js` - Search-and-replace file editing
+- `glob.js` - Find files matching glob patterns
+- `list_dir.js` - List directory contents with sizes
+
+**Code Search (1 tool):**
+- `grep.js` - Search file contents with regex support
+
+**System Operations (2 tools):**
+- `bash.js` - Execute shell commands with timeout support
+- `sleep.js` - Wait for specified duration
+
+**Task Management (5 tools):**
+- `task_create.js` - Create new async tasks
+- `task_status.js` - Check task status and output
+- `task_respond.js` - Respond to waiting tasks
+- `task_subscribe.js` - Subscribe to task completion
+
+**Memory Management (3 tools):**
+- `memory_read.js` - Read agent/project memory files
+- `memory_write.js` - Write to persistent memory
+- `memory_search.js` - Search across memory files
+
+**Planning (2 tools):**
+- `todo_read.js` - Read session TODO list
+- `todo_write.js` - Create/update structured TODOs
+
+**Web Operations (2 tools):**
+- `web_search.js` - Web search via DuckDuckGo
+- `web_fetch.js` - Fetch URL content (text/JSON/base64)
+
+**Utilities (1 tool):**
+- `tool_search.js` - Search available tools by name/description
+- `log_write.js` - Write progress entries to task log
+
+### Key Dependencies:
+- Most tools use `../infrastructure/database` for persistence
+- File tools use Node.js `fs` and `path` modules
+- Memory tools use `../utils/paths` for directory resolution

package/.veil/memory/agents/hello/MEMORY.md

@@ -0,0 +1,12 @@
+
+
+<!-- 2026-03-02 -->
+User's name is TestUser42.
+
+<!-- 2026-03-02 -->
+<!-- 2026-03-02 -->
+User's name is TestUser42.
+
+<!-- 2026-03-02 -->
+<!-- 2026-03-02 -->
+User's test codeword: **ZEBRA_KITE_77**

package/.veil/runtime.pid

@@ -0,0 +1 @@
+3672838

package/.veil/settings.json

@@ -0,0 +1,10 @@
+{
+  "port": 5050,
+  "maxIterations": 30,
+  "maxDurationSeconds": 120,
+  "permissions": {
+    "allow": ["*"],
+    "deny": [],
+    "ask": []
+  }
+}

package/PLAN/01-vision.md

@@ -0,0 +1,26 @@
+# 01 — Vision & Identity
+
+## What VeilCLI Is
+
+VeilCLI is an **enterprise autonomous agent runtime**. It runs locally on any machine, manages autonomous AI agents, and exposes a clean REST API for any consumer (web app, mobile app, CLI, another service) to interact with.
+
+Think: **rebuilt OpenClaw with Claude Code-level engineering quality** — but for autonomous enterprise agents, not a coding copilot.
+
+## What VeilCLI Is NOT
+
+- **Not a coding copilot** or IDE extension
+- **Not a cloud service** — runs locally, optional remote access via Ably
+- **Not a UI application** — REST API only, UIs built separately
+- **Not a personal/hobby tool** — enterprise-grade: permissions, hooks, audit logs, retention
+- **Not an LLM wrapper** — it's a full agent runtime with tool execution, context management, multi-agent orchestration
+
+## Core Philosophy
+
+- **Local-first, offline-capable** — no internet required for core operation
+- **API-first, UI-agnostic** — REST API is the sole interface; documentation is primary
+- **File-system for definitions** — agents, skills, config are files (versionable, portable)
+- **SQLite for runtime data** — sessions, messages, tasks (queryable, atomic, single-file)
+- **Claude Code ecosystem importable** — `veil import` converts CLAUDE.md, .claude/ settings, and skills to VeilCLI format
+- **Any LLM provider** — single `base_url` + `api_key` interface (OpenRouter, Anthropic, Ollama, etc.)
+- **Minimum custom work** — reuse existing libraries and standards, don't reinvent
+- **Forkable** — all branding/naming in `config.json`, can be forked and rebranded