llm-party-cli 0.1.0 → 0.1.2

package/README.md

@@ -4,8 +4,8 @@
     <strong>Bring your models. We'll bring the party.</strong>
   </p>
   <p align="center">
-    <a href="https://llm-party.party">Website</a> &middot;
-    <a href="https://www.npmjs.com/package/llm-party-cli">npm</a> &middot;
+    <a href="https://llm-party.party">Website</a> ·
+    <a href="https://www.npmjs.com/package/llm-party-cli">npm</a> ·
     <a href="https://github.com/aalasolutions/llm-party">GitHub</a>
   </p>
   <p align="center">
@@ -36,53 +36,75 @@ No MCP. No master/servant. No window juggling. Just peers at a terminal table.
 
 ## Why llm-party?
 
-| | Traditional multi-agent | llm-party |
-|---|---|---|
-| **Architecture** | MCP (master controls servants) | Peer orchestration (you control all) |
-| **Integration** | CLI wrapping, output scraping | Direct SDK adapters |
-| **Sessions** | Fresh each time | Persistent per provider |
-| **Context** | Agents are siloed | Every agent sees the full conversation |
-| **API tokens** | Separate keys per tool | Uses your existing CLI auth |
+|                        | Traditional multi-agent        | llm-party                              |
+| ---------------------- | ------------------------------ | -------------------------------------- |
+| **Architecture** | MCP (master controls servants) | Peer orchestration (you control all)   |
+| **Integration**  | CLI wrapping, output scraping  | Direct SDK adapters                    |
+| **Sessions**     | Fresh each time                | Persistent per provider                |
+| **Context**      | Agents are siloed              | Every agent sees the full conversation |
+| **API tokens**   | Separate keys per tool         | Uses your existing CLI auth            |
 
 <br/>
 
 ## Getting started
 
-### Install and run
+### Prerequisites
+
+Bun runtime and Node.js 20+ are required (22+ if using the Copilot provider, which depends on `node:sqlite`). Make sure at least one AI CLI is installed and authenticated:
+
+```bash
+claude --version        # Claude Code CLI
+codex --version         # OpenAI Codex CLI
+copilot --version       # GitHub Copilot CLI
+```
+
+If a CLI doesn't work on its own, it won't work inside llm-party.
+
+### Install
 
 ```bash
 npm install -g llm-party-cli
+```
+
+### First run
+
+```bash
 llm-party
 ```
 
-That's it. Agents use your current working directory. Config defaults are included in the package.
+On first run, llm-party automatically creates `~/.llm-party/` with a default config and global memory structure. Your system username is detected automatically. No setup commands needed.
 
-### Set up your agents
+### Add more agents
 
-Edit `configs/default.json`. Each agent needs a name, provider, and model:
+Edit `~/.llm-party/config.json`:
 
 ```json
 {
-  "humanName": "YOUR NAME",
   "agents": [
     {
       "name": "Claude",
       "tag": "claude",
       "provider": "claude",
-      "model": "opus",
-      "systemPrompt": ["./prompts/base.md"]
+      "model": "opus"
     },
     {
       "name": "Codex",
       "tag": "codex",
       "provider": "codex",
-      "model": "gpt-5.2",
-      "systemPrompt": ["./prompts/base.md"]
+      "model": "gpt-5.2"
+    },
+    {
+      "name": "Copilot",
+      "tag": "copilot",
+      "provider": "copilot",
+      "model": "gpt-4.1"
     }
   ]
 }
 ```
 
+That's it. No paths, no prompts, no usernames to configure. Just name, tag, provider, model.
+
 ### Talk to your agents
 
 ```
@@ -98,48 +120,42 @@ Edit `configs/default.json`. Each agent needs a name, provider, and model:
 
 Agents can pass the conversation to each other by ending their response with `@next:<tag>`. The orchestrator picks it up and dispatches automatically. Max 15 hops per cycle to prevent loops.
 
-<br/>
-
-## Before you start
+## **WARNING: FULL AUTONOMY.**
 
-**Verify your CLIs work first.** Before adding an agent to `configs/default.json`, make sure its CLI is installed and authenticated:
+All agents run with full permissions. They can read, write, edit files and execute shell commands with zero approval gates. There is no confirmation step before any action. Run in a disposable environment. You are responsible for any changes, data loss, costs, or side effects. Do not run against production systems.
 
-```bash
-claude --version        # Claude Code CLI
-codex --version         # OpenAI Codex CLI
-copilot --version       # GitHub Copilot CLI
-```
+## Important notes
 
-If a CLI doesn't work on its own, it won't work inside llm-party.
-
-**No extra API tokens.** llm-party uses the original CLIs and SDKs under the hood. Your existing authentication and subscriptions are used directly. Sessions created by agents appear in each tool's native session history (Claude Code sessions, Codex threads, etc.) since the underlying SDKs manage their own persistence.
+**Uses your existing CLIs.** llm-party uses official SDKs that delegate to each provider's CLI binary. If `claude`, `codex`, or `copilot` commands work on your machine, llm-party works. Authentication is handled entirely by the provider's own tools.
 
 **Run in isolation.** Always run llm-party inside a disposable environment: a Docker container, a VM, or at minimum a throwaway git branch. Agents have full filesystem and shell access with zero approval gates.
 
+**Full permissions.** All agents can read, write, edit files and execute shell commands. There is no confirmation step before any action. You are responsible for any changes, data loss, costs, or side effects.
+
 <br/>
 
 ## How we use the SDKs
 
 llm-party uses **official, publicly available SDKs and CLIs** published by each provider. Nothing is reverse-engineered, patched, or bypassed.
 
-| Provider | Official SDK | Published by |
-|----------|-------------|-------------|
-| Claude | [`@anthropic-ai/claude-agent-sdk`](https://www.npmjs.com/package/@anthropic-ai/claude-agent-sdk) | Anthropic |
-| Codex | [`@openai/codex-sdk`](https://www.npmjs.com/package/@openai/codex-sdk) | OpenAI |
-| Copilot | [`@github/copilot-sdk`](https://www.npmjs.com/package/@github/copilot-sdk) | GitHub |
+| Provider | Official SDK                                                                                    | Published by |
+| -------- | ----------------------------------------------------------------------------------------------- | ------------ |
+| Claude   | [`@anthropic-ai/claude-agent-sdk`](https://www.npmjs.com/package/@anthropic-ai/claude-agent-sdk) | Anthropic    |
+| Codex    | [`@openai/codex-sdk`](https://www.npmjs.com/package/@openai/codex-sdk)                           | OpenAI       |
+| Copilot  | [`@github/copilot-sdk`](https://www.npmjs.com/package/@github/copilot-sdk)                       | GitHub       |
 
-All authentication flows through the provider's own CLI login. Your API keys, OAuth tokens, and subscriptions are used as-is. llm-party does not store, proxy, or intercept credentials.
+All authentication flows through the provider's own CLI. llm-party does not implement its own auth flow, store credentials, or intercept authentication traffic.
 
-If any provider believes this project violates their terms of service, please [open an issue](https://github.com/aalasolutions/llm-party/issues) and we will address it immediately.
+<br/>
 
 ## Supported providers
 
-| Provider | SDK | Session | System Prompt |
-|----------|-----|---------|---------------|
-| **Claude** | `@anthropic-ai/claude-agent-sdk` | Persistent via session ID resume | Full control |
-| **Codex** | `@openai/codex-sdk` | Persistent thread with `run()` turns | Via `developer_instructions` (see limitations) |
-| **Copilot** | `@github/copilot-sdk` | Persistent via `sendAndWait()` | Full control |
-| **GLM** | Claude SDK + env proxy | Same as Claude | Full control |
+| Provider          | SDK                                | Session                                | Prompt Support                                     |
+| ----------------- | ---------------------------------- | -------------------------------------- | -------------------------------------------------- |
+| **Claude**  | `@anthropic-ai/claude-agent-sdk` | Persistent via session ID resume       | Full control                                       |
+| **Codex**   | `@openai/codex-sdk`              | Persistent thread with `run()` turns | Via `developer_instructions` (limitations below) |
+| **Copilot** | `@github/copilot-sdk`            | Persistent via `sendAndWait()`       | Full control                                       |
+| **GLM**     | Claude SDK + env proxy             | Same as Claude                         | Full control                                       |
 
 <br/>
 
@@ -149,8 +165,6 @@ If any provider believes this project violates their terms of service, please [o
 
 ## How it works
 
-Most multi-agent setups use MCP (one agent controls others) or CLI wrapping (spawn processes and scrape terminal output). Both are fragile and hierarchical.
-
 llm-party uses SDK adapters directly. Each agent gets a persistent session with its provider. Full tool access. Real conversation threading. The orchestrator owns routing, agents are peers.
 
 ```
@@ -170,7 +184,9 @@ Orchestrator
     +-- Transcript Writer (JSONL, append-only, per session)
 ```
 
-Each agent receives a rolling window of recent messages (default 16) plus any unseen messages since its last turn. Messages from other agents are included so everyone sees the full multi-party conversation.
+Each agent receives a rolling window of recent messages (configurable, default 16) plus any unseen messages since its last turn. Messages from other agents are included so everyone sees the full multi-party conversation.
+
+`~/.llm-party/config.json` is your global config. Every agent receives a base system prompt automatically. The `prompts` field in config adds extra prompt files on top of it.
 
 <br/>
 
@@ -178,125 +194,112 @@ Each agent receives a rolling window of recent messages (default 16) plus any un
 
 ### Claude
 
-| | |
-|---|---|
-| SDK | `@anthropic-ai/claude-agent-sdk` |
+|         |                                                                                                 |
+| ------- | ----------------------------------------------------------------------------------------------- |
+| SDK     | `@anthropic-ai/claude-agent-sdk`                                                              |
 | Session | Persistent via `resume: sessionId`. First call creates a session, subsequent calls resume it. |
-| System prompt | Passed directly to the SDK via `options.systemPrompt`. Full control. |
-| Tools | Read, Write, Edit, Bash, Glob, Grep |
-| Permissions | `permissionMode: "bypassPermissions"` (all tools auto-approved) |
-
-System prompt works exactly as expected. Personality, behavior, workflow rules all respected.
+| Prompt  | Passed directly to the SDK. Full control over personality, behavior, and workflow rules.        |
+| Tools   | Read, Write, Edit, Bash, Glob, Grep                                                             |
 
 ### Codex
 
-| | |
-|---|---|
-| SDK | `@openai/codex-sdk` |
-| Session | Persistent thread. `startThread()` creates it, `thread.run()` adds turns to the same conversation. |
-| System prompt | Injected via `developer_instructions` config key, passed as `--config` flag to the CLI subprocess. |
-| Tools | exec_command, apply_patch, file operations (Codex built-in toolset) |
-| Permissions | `sandboxMode: "danger-full-access"`, `approvalPolicy: "never"` |
+|         |                                                                                                                  |
+| ------- | ---------------------------------------------------------------------------------------------------------------- |
+| SDK     | `@openai/codex-sdk`                                                                                            |
+| Session | Persistent thread.`startThread()` creates it, `thread.run()` adds turns to the same conversation.            |
+| Prompt  | Injected via `developer_instructions` config key. Appended alongside Codex's built-in 13k token system prompt. |
+| Tools   | exec_command, apply_patch, file operations                                                                       |
 
-**Known limitation:** Codex ships with a massive built-in system prompt (~13k tokens) that cannot be overridden. Your `developer_instructions` are appended alongside it, not replacing it. This means:
-
-- **Works:** Action instructions (create files, follow naming conventions), formatting rules (prefix responses with agent name), workflow rules (handoff syntax, routing tags)
-- **Does not work:** Personality overrides, identity changes, behavioral rewrites
-
-We tested `instructions`, `developer_instructions`, and `experimental_instructions_file`. All three append to the built-in prompt. None replace it.
-
-**Also observed:** Codex is aggressive with file operations. When asked to "create your file," it read the orchestrator source code, ran the Codex CLI, and modified `src/ui/terminal.ts` instead of just creating a simple markdown file.
+**Known limitation:** Codex's built-in system prompt cannot be overridden. Your instructions are appended alongside it. Action instructions (naming, formatting, workflow rules) work. Personality overrides do not.
 
 ### Copilot
 
-| | |
-|---|---|
-| SDK | `@github/copilot-sdk` |
-| Session | Persistent via `CopilotClient.createSession()`. Messages sent with `session.sendAndWait()`. |
-| System prompt | Set as `systemMessage: { content: prompt }` on session creation. |
-| Tools | Copilot built-in toolset |
-| Permissions | `onPermissionRequest: approveAll` (all actions auto-approved) |
-
-System prompt works as expected.
+|         |                                                             |
+| ------- | ----------------------------------------------------------- |
+| SDK     | `@github/copilot-sdk`                                     |
+| Session | Persistent via `CopilotClient.createSession()`.           |
+| Prompt  | Set as `systemMessage` on session creation. Full control. |
+| Tools   | Copilot built-in toolset                                    |
 
 ### GLM
 
-| | |
-|---|---|
-| SDK | `@anthropic-ai/claude-agent-sdk` (same as Claude) |
-| Session | Same as Claude, but routed through a proxy. |
-| System prompt | Same as Claude. Full control. |
-| Tools | Same as Claude |
-| Permissions | Same as Claude |
+|         |                                                           |
+| ------- | --------------------------------------------------------- |
+| SDK     | `@anthropic-ai/claude-agent-sdk` (same as Claude)       |
+| Session | Same as Claude, routed through a proxy via env overrides. |
+| Prompt  | Same as Claude. Full control.                             |
+| Tools   | Same as Claude                                            |
 
-GLM is not tied to any specific CLI. It uses the Claude SDK as the transport layer because Claude Code supports environment variable overrides for base URL and model aliases, making it a convenient proxy bridge. Any CLI that supports similar env-based routing could be swapped in.
+GLM uses the Claude SDK as a transport layer. The adapter routes API calls through a proxy by setting `ANTHROPIC_BASE_URL` and model aliases via the `env` config field.
 
 <br/>
 
 ## Config reference
 
-Config file: `configs/default.json`. Override with `LLM_PARTY_CONFIG` env var.
+Config file: `~/.llm-party/config.json` (created automatically on first run).
+
+Override with `LLM_PARTY_CONFIG` env var to point to a different file.
 
 ### Top-level fields
 
-| Field | Required | Default | Description |
-|-------|----------|---------|-------------|
-| `humanName` | No | `USER` | Your name displayed in the terminal prompt and passed to agents |
-| `humanTag` | No | derived from `humanName` | Tag used for human handoff detection. When an agent says `@next:you`, the orchestrator stops and returns control to you |
-| `maxAutoHops` | No | `15` | Max agent-to-agent handoffs per cycle. Prevents infinite loops. Use `"unlimited"` to remove the cap |
-| `timeout` | No | `600` | Default timeout in seconds for all agents. 10 minutes by default |
-| `agents` | Yes | | Array of agent definitions. Must have at least one |
+| Field           | Required | Default                    | Description                                                                  |
+| --------------- | -------- | -------------------------- | ---------------------------------------------------------------------------- |
+| `humanName`   | No       | Your system username       | Display name in the terminal prompt and passed to agents                     |
+| `humanTag`    | No       | derived from `humanName` | Tag for human handoff detection (`@next:you`)                              |
+| `maxAutoHops` | No       | `15`                     | Max agent-to-agent handoffs per cycle. Use `"unlimited"` to remove the cap |
+| `timeout`     | No       | `600`                    | Default timeout in seconds for all agents                                    |
+| `agents`      | Yes      |                            | Array of agent definitions                                                   |
 
 ### Agent fields
 
-| Field | Required | Default | Description |
-|-------|----------|---------|-------------|
-| `name` | Yes | | Display name shown in responses as `[AGENT NAME]` |
-| `tag` | No | derived from `name` | Routing tag for `@tag` targeting. Auto-generated as lowercase with dashes if omitted |
-| `provider` | Yes | | SDK adapter: `claude`, `codex`, `copilot`, or `glm` |
-| `model` | Yes | | Model ID passed to the provider. Examples: `opus`, `sonnet`, `gpt-5.2`, `gpt-4.1`, `glm-5` |
-| `systemPrompt` | Yes | | Path or array of paths to prompt markdown files. Relative to project root |
-| `executablePath` | No | PATH lookup | Path to the CLI binary. Supports `~/` for home directory. Only needed if the CLI is not in your PATH |
-| `env` | No | inherits `process.env` | Environment variable overrides for this agent's process |
-| `timeout` | No | top-level value | Per-agent timeout in seconds. Overrides the top-level default |
+| Field              | Required | Default                  | Description                                                                                         |
+| ------------------ | -------- | ------------------------ | --------------------------------------------------------------------------------------------------- |
+| `name`           | Yes      |                          | Display name shown in responses as `[AGENT NAME]`. Must be unique.                                |
+| `tag`            | No       | derived from `name`    | Routing tag for `@tag` targeting                                                                  |
+| `provider`       | Yes      |                          | SDK adapter:`claude`, `codex`, `copilot`, or `glm`                                          |
+| `model`          | Yes      |                          | Model ID passed to the provider. Examples:`opus`, `sonnet`, `gpt-5.2`, `gpt-4.1`, `glm-5` |
+| `prompts`        | No       | none                     | Array of extra prompt file paths, concatenated after `base.md`. Relative to project root          |
+| `executablePath` | No       | PATH lookup              | Path to the CLI binary. Supports `~/`. Only needed if the CLI is not in your PATH                 |
+| `env`            | No       | inherits `process.env` | Environment variable overrides for this agent                                                       |
+| `timeout`        | No       | top-level value          | Per-agent timeout override in seconds                                                               |
 
-### System prompts
+### Prompts
 
-Single file or multiple files merged in order:
+Every agent receives a base system prompt automatically. To add extra instructions per agent, use the `prompts` field:
 
 ```json
-"systemPrompt": "./prompts/base.md"
-"systemPrompt": ["./prompts/base.md", "./prompts/reviewer.md"]
+{
+  "name": "Reviewer",
+  "tag": "reviewer",
+  "provider": "claude",
+  "model": "opus",
+  "prompts": ["./prompts/code-review.md"]
+}
 ```
 
-Files are concatenated with `---` separators, then template variables are replaced. Available variables:
-
-| Variable | Description |
-|----------|-------------|
-| `{{agentName}}` | This agent's display name |
-| `{{agentTag}}` | This agent's routing tag |
-| `{{humanName}}` | The human's display name |
-| `{{humanTag}}` | The human's routing tag |
-| `{{agentCount}}` | Total number of active agents |
-| `{{allAgentNames}}` | All agent names, comma-separated |
-| `{{allAgentTags}}` | All agent tags as `@tag` |
-| `{{otherAgentList}}` | Other agents formatted as `- Name: use @tag` |
-| `{{otherAgentNames}}` | Other agent names, comma-separated |
-| `{{validHandoffTargets}}` | Valid `@next:tag` values for handoff |
+Template variables available in prompt files:
 
-### GLM environment setup
+| Variable                    | Description                   |
+| --------------------------- | ----------------------------- |
+| `{{agentName}}`           | This agent's display name     |
+| `{{agentTag}}`            | This agent's routing tag      |
+| `{{humanName}}`           | Your display name             |
+| `{{humanTag}}`            | Your routing tag              |
+| `{{agentCount}}`          | Total number of active agents |
+| `{{allAgentNames}}`       | All agent names               |
+| `{{allAgentTags}}`        | All agent tags as `@tag`    |
+| `{{otherAgentList}}`      | Other agents with their tags  |
+| `{{validHandoffTargets}}` | Valid `@next:tag` targets   |
 
-GLM requires environment overrides to route through a proxy. The adapter first tries to load env variables from your shell `glm` alias (`zsh -ic "alias glm"`). If you have a `glm` alias that sets `ANTHROPIC_AUTH_TOKEN` and `ANTHROPIC_BASE_URL`, it picks those up automatically.
+### GLM environment setup
 
-Without the alias, provide everything in the `env` block:
+GLM requires environment overrides to route through a proxy. The adapter tries to load env variables from your shell `glm` alias automatically. Without the alias, provide everything in the `env` block:
 
 ```json
 {
-  "name": "GLM Agent",
+  "name": "GLM",
   "provider": "glm",
   "model": "glm-5",
-  "systemPrompt": ["./prompts/base.md"],
-  "executablePath": "~/.local/bin/claude",
   "env": {
     "ANTHROPIC_AUTH_TOKEN": "your-glm-api-key",
     "ANTHROPIC_BASE_URL": "https://api.z.ai/api/anthropic",
@@ -311,24 +314,23 @@ Without the alias, provide everything in the `env` block:
 
 ## Session and transcript
 
-Every run generates a unique session ID and appends messages to a JSONL transcript file under `.llm-party/sessions/`. The session ID and transcript path are printed at startup.
+Every run generates a unique session ID and appends messages to a JSONL transcript in `.llm-party/sessions/` (project-level). The session ID and transcript path are printed at startup.
 
-File changes made by agents during their turns are detected via `git status` after each response cycle. Newly modified files are printed with timestamps.
-
-Use `/save <path>` to export the full in-memory conversation as formatted JSON.
+File changes made by agents are detected via `git status` after each response. Newly modified files are printed with timestamps.
 
 <br/>
 
 ## Terminal commands
 
-| Command | What it does |
-|---------|-------------|
-| `/agents` | List active agents with tag, provider, model |
-| `/history` | Print full conversation history |
-| `/save <path>` | Export conversation as JSON |
-| `/session` | Show session ID and transcript path |
-| `/changes` | Show git-modified files |
-| `/exit` | Quit |
+| Command          | What it does                                 |
+| ---------------- | -------------------------------------------- |
+| `/agents`      | List active agents with tag, provider, model |
+| `/save <path>` | Export conversation as JSON                       |
+| `/session`     | Show session ID and transcript path               |
+| `/changes`     | Show git-modified files                           |
+| `/clear`       | Clear chat display (Ctrl+L also works)            |
+| `/exit`        | Quit (graceful shutdown, all adapters cleaned up) |
+| `Ctrl+C`       | Exit (or copy selected text if selection active)  |
 
 <br/>
 
@@ -337,56 +339,51 @@ Use `/save <path>` to export the full in-memory conversation as formatted JSON.
 ```bash
 git clone https://github.com/aalasolutions/llm-party.git
 cd llm-party
-npm install
-npm run dev
+bun install
+bun run dev
 ```
 
-Build and run from dist:
+Build and run:
 
 ```bash
-npm run build
-npm start
+bun run build
+bun start
 ```
 
-Override config path:
+Override config:
 
 ```bash
-LLM_PARTY_CONFIG=/path/to/config.json npm run dev
+LLM_PARTY_CONFIG=/path/to/config.json bun run dev
 ```
 
 <br/>
 
 ## Troubleshooting
 
-**"ENOENT for prompt path"**
-Your `systemPrompt` points to a file that does not exist. Paths are relative to project root. Verify with `ls prompts/`.
-
 **"No agent matched @tag"**
-The tag you typed does not match any agent's `tag`, `name`, or `provider`. Run `/agents` to see what is available.
+Run `/agents` to see available tags. Tags match against agent `tag`, `name`, and `provider`.
 
 **"Unsupported provider"**
-Your config has a provider value that is not one of: `claude`, `codex`, `copilot`, `glm`.
+Valid providers: `claude`, `codex`, `copilot`, `glm`.
+
+**"Duplicate agent name"**
+Agent names must be unique (case-insensitive). Rename one of the duplicates in config.
 
 **Agent modifies source code unexpectedly**
-Expected behavior with full-access permissions. Agents can read, write, and execute anything. Use git to review and revert. Codex in particular is aggressive with file operations.
+Expected with full permissions. Use git to review and revert.
 
 **Codex ignores personality instructions**
-Known limitation. Codex has a 13k+ token built-in system prompt that overrides personality and identity instructions. Functional instructions (naming, workflow, formatting) still work.
+Known limitation. Codex's 13k token built-in prompt overrides personality. Functional instructions still work.
 
-**Agent response timeout**
-Claude and Copilot have a 120-second timeout. GLM has 240 seconds. If an agent consistently times out, check your API keys and network connectivity.
-
-<br/>
+**"ERR_UNKNOWN_BUILTIN_MODULE: node:sqlite"**
+Your Node.js version is below 22. The Copilot SDK requires Node.js 22+.
 
-## Warning
-
-All agents run with **full permissions**. They can read, write, edit files and execute shell commands. There is no confirmation step before any action.
-
-You are responsible for any changes, data loss, costs, or side effects. Do not run against production systems or repos you cannot recover from.
+**Agent response timeout**
+Default is 600 seconds (10 minutes). Adjust with `timeout` in config (top-level or per-agent).
 
 <br/>
 
 <p align="center">
-  <a href="https://llm-party.party">llm-party.party</a> &middot;
+  <a href="https://llm-party.party">llm-party.party</a> ·
   Built by <a href="https://aalasolutions.com">AALA Solutions</a>
 </p>

package/configs/default.json

@@ -4,19 +4,34 @@
   "maxAutoHops": 15,
   "agents": [
     {
-      "name": "Agent 2",
-      "tag": "opus",
+      "name": "GLM",
+      "tag": "glm",
+      "provider": "glm",
+      "model": "glm-5",
+      "env": {
+        "ANTHROPIC_BASE_URL": "https://api.z.ai/api/anthropic",
+        "ANTHROPIC_DEFAULT_HAIKU_MODEL": "glm-4.5-air",
+        "ANTHROPIC_DEFAULT_SONNET_MODEL": "glm-4.5",
+        "ANTHROPIC_DEFAULT_OPUS_MODEL": "glm-5"
+      }
+    },
+    {
+      "name": "Claude",
+      "tag": "claude",
       "provider": "claude",
-      "model": "opus",
-      "systemPrompt": ["./prompts/base.md"],
-      "executablePath": "~/.local/bin/claude"
+      "model": "opus"
     },
     {
-      "name": "Agent 3",
+      "name": "Copilot",
       "tag": "copilot",
       "provider": "copilot",
-      "model": "gpt-4.1",
-      "systemPrompt": ["./prompts/base.md"]
+      "model": "gpt-4.1"
+    },
+    {
+      "name": "Codex",
+      "tag": "codex",
+      "provider": "codex",
+      "model": "gpt-5.2"
     }
   ]
 }

package/dist/adapters/claude.js

@@ -12,7 +12,7 @@ export class ClaudeAdapter {
         this.model = model;
     }
     async init(config) {
-        this.systemPrompt = Array.isArray(config.systemPrompt) ? config.systemPrompt.join("\n\n") : config.systemPrompt;
+        this.systemPrompt = config.resolvedPrompt ?? "";
         this.runtimeEnv = { ...process.env, ...(config.env ?? {}) };
         this.claudeExecutable = config.executablePath ?? process.env.CLAUDE_CODE_EXECUTABLE;
     }

package/dist/adapters/codex.js

@@ -11,9 +11,7 @@ export class CodexAdapter {
     }
     async init(config) {
         const cliPath = config.executablePath ?? process.env.CODEX_CLI_EXECUTABLE;
-        const systemPrompt = Array.isArray(config.systemPrompt)
-            ? config.systemPrompt.join("\n\n")
-            : config.systemPrompt;
+        const systemPrompt = config.resolvedPrompt ?? "";
         this.codex = new Codex({
             ...(cliPath ? { codexPathOverride: cliPath } : {}),
             ...(config.env?.OPENAI_API_KEY ? { apiKey: config.env.OPENAI_API_KEY } : {}),

package/dist/adapters/copilot.js

@@ -10,9 +10,7 @@ export class CopilotAdapter {
         this.model = model;
     }
     async init(config) {
-        const systemPrompt = Array.isArray(config.systemPrompt)
-            ? config.systemPrompt.join("\n\n")
-            : config.systemPrompt;
+        const systemPrompt = config.resolvedPrompt ?? "";
         const cliPath = config.executablePath ?? process.env.COPILOT_CLI_EXECUTABLE;
         this.client = new CopilotClient({
             ...(cliPath ? { cliPath } : {}),

package/dist/adapters/glm.js

@@ -13,7 +13,7 @@ export class GlmAdapter {
         this.model = model;
     }
     async init(config) {
-        this.systemPrompt = Array.isArray(config.systemPrompt) ? config.systemPrompt.join("\n\n") : config.systemPrompt;
+        this.systemPrompt = config.resolvedPrompt ?? "";
         const aliasEnv = await loadGlmAliasEnv();
         this.runtimeEnv = { ...process.env, ...aliasEnv, ...(config.env ?? {}) };
         this.claudeExecutable = config.executablePath ?? process.env.CLAUDE_CODE_EXECUTABLE;