Instruction file imported from mandersogit/agent-chain (
.cursor/rules/agent-chain.mdc). Copyright stays with the author.
agent-chain
Multi-agent chain orchestration CLI. Runs sequences of AI coding agents (codex-cli, claude-code) with structured handoffs, collecting telemetry and producing completion reports.
Project Structure
agent-chain/
├── pyproject.toml
├── Makefile
├── src/agent_chain/ # Source package (src layout)
│ ├── __init__.py
│ ├── __main__.py
│ ├── cli.py # click CLI entry point
│ ├── chain.py # ChainDefinition, ChainLoader
│ ├── runner.py # ChainRunner
│ ├── report.py # CompletionReport, ReportWriter
│ ├── telemetry.py # TelemetryRecord, aggregation
│ ├── variables.py # Template variable resolution
│ ├── types.py # Shared types (StepConfig, AgentResult)
│ └── backends/ # Agent backend implementations
├── tests/
├── scripts/commit-helper.py # Polyglot commit plan executor
└── workflow/ # Design docs and working artifacts
Development Commands
make test # Run all tests
make test-cov # Run tests with coverage
make lint # Lint with ruff
make typecheck # Run mypy + pyright
make format # Format with ruff
make all # lint + typecheck + test
make clean # Remove build artifacts
make install # pip install -e '.[dev]' into local.venvPython Conventions
- Python 3.11+ only
- Qualified imports everywhere (see coding-standards.mdc)
- External imports: underscore prefix (
import click as _click) - Internal imports: no prefix (
import agent_chain.chain as chain) - Modern type hints:
list[str],X | None, notyping.List
Design Document
The full design lives at workflow/2026-02-25-agent-chain-tool-design.md.