athola/claude-night-market/minimax-delegation

MiniMax CLI Delegation

Overview

This skill implements conjure:delegation-core for the official MiniMax CLI, mmx. It provides MiniMax-specific authentication, quota management, and command construction.

For shared delegation patterns, see Skill(conjure:delegation-core).

When To Use

• After Skill(conjure:delegation-core) determines MiniMax is suitable
• When you need Flash or M3's large context window (1M tokens)
• For batch processing, summarization, or multi-file analysis
• If the mmx CLI is installed and authenticated

When NOT To Use

• Choosing which provider to delegate to (use conjure:delegation-core)
• Another provider was selected (use that provider's delegation skill)

Prerequisites

Installation:

Install the official CLI, published as mmx-cli and maintained by MiniMax at https://github.com/MiniMax-AI/cli. It requires Node.js 18 or later and installs a binary named mmx.

npm install -g mmx-cli
mmx --version

Install mmx-cli by that exact name. Unrelated third-party packages publish a binary called minimax, and this skill never invokes that name.

Authentication:

Credentials are managed by the CLI, not by an environment variable. mmx stores them in ~/.mmx/config.json.

mmx auth login                      # interactive: OAuth or an API key
mmx auth login --api-key sk-xxxxx   # non-interactive, for CI
mmx auth status                     # canonical readiness check

mmx auth status is what delegation_executor.py --verify runs. A successful mmx --version alone does not mean the CLI is authenticated.

Regions: MiniMax serves a global region (api.minimax.io) and a mainland China region (api.minimaxi.com). API keys are issued per region and are not interchangeable. API-key login detects the matching region by probing both, so no manual base URL is normally needed. Set the region explicitly with mmx config set --key region --value cn or the MINIMAX_REGION environment variable.

Quick Start

Using Shared Delegation Executor

# Basic file analysis
python scripts/delegation_executor.py minimax "Analyze this code" --files src/main.py
 
# With specific model
python scripts/delegation_executor.py minimax "Summarize" --files src/ --model MiniMax-M3
 
# Verify the CLI is installed and authenticated
python scripts/delegation_executor.py minimax --verify

The executor reads the requested files and inlines their contents into the prompt, because mmx has no @path context syntax. Inlined context is capped at 96 KiB to stay inside the operating system limit on a single argument, and any remainder is reported in the prompt and the log.

Direct CLI Usage

# Basic command
mmx text chat --model MiniMax-M3.1-Flash --message "Analyze this code"
 
# Specific model
mmx text chat --model MiniMax-M3 --message "Summarize this design"
 
# JSON output
mmx text chat --message "Extract the function names" --output json
 
# File contents must be passed in, not referenced
mmx text chat --message "Review this: $(cat src/main.py)"

Save Output

mmx text chat --message "..." > delegations/minimax/$(date +%Y%m%d_%H%M%S).md

Smart Delegation

The shared delegation executor can auto-select the best service:

uv run python scripts/delegation_executor.py auto "Analyze large codebase" \
  --files src/

MiniMax-Specific Details

For MiniMax-specific models, CLI options, cost reference, and troubleshooting, see modules/minimax-specifics.md.

Exit Criteria

• [ ] mmx --version and mmx auth status both exit 0 before any task is

delegated. A missing installation or failed authentication is reported and stops execution.

• [ ] Every spawned command starts with the mmx binary. No command in this

skill or in Delegator.SERVICES["minimax"] invokes a binary named minimax.

• [ ] File context reaches the model as inlined contents. No @path reference

appears in a prompt sent to mmx.

• [ ] The delegated task output is saved to

delegations/minimax/YYYYMMDD_HHMMSS.md (timestamp format matching the Quick Start example), and that file exists on disk after delegation.

• [ ] If conjure:delegation-core selected a different provider, this skill is

not invoked. MiniMax delegation only runs when delegation-core routes to it.