@mark-heartflow

claude-code-custom-model-proxy bridges anthropic messages api and openai chat completions format via a local proxy server, enabling claude code to work with third-party model providers like minimax.

claude-statusline-skill provides a bash-based statusline for claude code displaying git branch and model name, with placeholder architecture for context usage, tokens, cost, duration, and line changes. most metrics unimplemented.

self-improving-agent-pro-v1 advertises a framework for autonomous agent self-correction and continuous learning via memory persistence and decision verification, but lacks actionable procedure, concrete triggers, and failure mode analysis.

self-improving-agent-pro-v2 packages decision verification, memory routing, and q-learning into an agent framework. largely conceptual with minimal actionable procedures or concrete trigger definitions.

still-growing attempts to reframe parenting problems as parent trauma responses rather than child defects, grounded in philosophy and research citations. execution collapses into prose poetry and scattered prompts without actionable procedure.