Add Ralph Harness — guarded loop scaffold for coding agents

[rxdt]

Project: Ralph Harness
URL: https://github.com/rxdt/py_ralph_frame
PyPI: https://pypi.org/project/ralph-harness/

Suggested sections:

• Terminal and CLI Agents
• Lightweight / Minimalist
• Observability and Evaluation
• AI Safety and Guardrails

Short description:
Ralph Harness is a tiny Python scaffold for running Claude Code, Codex CLI, Gemini CLI, and similar coding agents in guarded local loops. It uses repo-local specs, fresh-context iterations, hard caps, a git-hook gate, CI verification, and coverage gates.

Why it fits:

• Built for developers already using coding agents who want local proof instead of a large orchestrator.
• No hosted backend, no token service, no agent lock-in.
• Targets a real coding-agent failure mode: drift, shallow tests, and progress claims without repo-level verification.
• Installable with uvx ralph-harness demo and uvx ralph-harness install <project>.

Suggested entry:
| Ralph Harness | Python scaffold for guarded Claude Code/Codex/Gemini loops with repo-local specs, fresh-context runs, git-hook gate, CI verification, and coverage gates. | Free (OSS) |