glm-plan-usage: Instructions file for Claude Code

glm-plan-usage

A Claude Code status bar plugin that displays GLM (ZHIPU/ZAI) coding plan usage statistics. Shows token usage percentage, countdown to reset, and MCP usage in real-time.

Tech Stack

Core: Rust (edition 2021) + clap + ureq + serde + toml; Packaging: npm

Commands

# Build & Run
cargo build --release          # Build release binary
./build-all.sh                 # Cross-platform build (7 targets)

# Lint & Format
cargo fmt                      # Format code
cargo clippy -- -D warnings    # Run linter

# Testing
cargo test                     # Run all tests (unit + e2e)
cargo test --test e2e          # Run e2e tests only
cargo test --lib               # Run unit tests only

# CLI
glm-plan-usage init            # Initialize config
glm-plan-usage print           # Print current config
glm-plan-usage check           # Validate config

Project Structure

For detailed module organization, see specs/directory-structure.md.

glm-plan-usage/
├── src/
│   ├── main.rs              # Entry point, stdin parsing, CLI handling
│   ├── cli.rs               # CLI argument definitions (clap derive)
│   ├── lib.rs               # Library interface
│   ├── config/              # Configuration loading and types
│   ├── api/                 # API client, cache, response types
│   └── core/                # Status line generation and segments
├── e2e/                    # End-to-end CLI tests (assert_cmd + httpmock)
│   ├── main.rs              # Test entry point
│   ├── helpers.rs           # Shared test utilities
│   ├── fixtures/            # JSON fixtures (stdin input, API responses)
│   └── tests/               # Test modules by workflow
├── npm/                     # NPM packaging (main + platform binaries)
├── specs/                   # Design specifications
└── build-all.sh             # Cross-platform build script

Standards

MUST FOLLOW THESE RULES, NO EXCEPTIONS

• Graceful degradation: the plugin must never cause Claude Code to fail. All segments return Option<SegmentData>, never panic.
• All config fields must have #[serde(default)] to allow adding new fields without breaking existing configs.
• Silent by default: no stderr output unless --verbose is set. API failures return None, not errors.
• Only add meaningful comments explaining why (not what) something is done.