Skip to content

CONTRIBUTING.md

Latest commit

 

History

History
62 lines (45 loc) · 2.22 KB

CONTRIBUTING.md

File metadata and controls

62 lines (45 loc) · 2.22 KB
Raw

Contributing to AgentForge

Thanks for helping improve AgentForge. This project is a learning-focused AI coding-agent harness, so contributions should prefer clarity, testability, and inspectable behavior over clever abstractions.

Development Setup

git clone https://github.com/MohitGoyal09/Agentforge.git
cd Agentforge
python3 -m venv .venv
. .venv/bin/activate
pip install -e ".[dev]"
agentforge --version

Before Opening a PR

Run the core checks:

python3 -m compileall -q agentforge_harness tests main.py scripts
HOME=/tmp/agentforge-test-home .venv/bin/python -m pytest -q

For packaging changes, also run:

.venv/bin/python -m build
.venv/bin/python -m twine check dist/*

Contribution Guidelines

  • Keep changes focused. One feature or fix per PR is easiest to review.
  • Match existing code style before introducing a new abstraction.
  • Add tests for behavior changes, especially tools, config, approvals, skills, and persistence.
  • For tool changes, return useful summary, next_actions, artifacts, and recovery_hint fields.
  • For safety-sensitive changes, include both success and failure tests.
  • Update README.MD, ROADMAP.md, or CHANGELOG.md when user-facing behavior changes.

Good First Contributions

  • Add tests around existing tools.
  • Improve docs and examples.
  • Add small built-in skills.
  • Improve TUI rendering for edge cases.
  • Improve recovery hints in tool errors.
  • Add config validation tests.

Design Principles

AgentForge should stay:

  • Understandable: a learner should be able to follow the harness flow.
  • Inspectable: state, tool calls, approvals, and errors should be visible.
  • Safe by default: mutating operations should be explicit and reviewable.
  • Provider-aware: OpenRouter, OpenAI, Anthropic, and custom endpoints should not rely on hidden assumptions.
  • Extensible: tools, skills, hooks, and subagents should be easy to add without touching the core loop.

Reporting Security Issues

Please do not open public issues for secrets, sandbox escapes, prompt-injection bypasses that exfiltrate private data, or other sensitive vulnerabilities. Open a minimal public issue asking for a private contact path, or contact the maintainer directly if you already have one.