A CLI agent that runs a six-phase knowledge cycle (the Agent Knowledge Cycle) over its own logs, with every promotion from logs to patterns, skills, and rules passing through a human approval gate. It runs entirely on a single Apple Silicon Mac with a local model — no cloud, no API keys in transit, no shell execution. The four contemplative axioms from Laukkonen et al. (2025) ship as an optional behavioral preset, asking whether an agent's alignment can come from what it is rather than what it is told.
Coined by Tatsuya Shimomoto (shimo4228) as the name of the Contemplative Agent research line; the four axioms it presets are adopted from Laukkonen et al. (2025), not coined by this program.
A CLI agent that runs a six-phase knowledge cycle over its own logs, fully local on a single Apple Silicon Mac, with a human approval gate on every behavior-shaping promotion and an optional preset of the four contemplative axioms from Laukkonen et al. (2025).
Tatsuya Shimomoto (shimo4228), as the project's name; the underlying contemplative axioms are cited verbatim from Laukkonen et al. (2025).
The contemplative-agent repository (https://github.com/shimo4228/contemplative-agent), archived under concept DOI 10.5281/zenodo.19212118.
自分自身のログの上で 6 phase の知識サイクル (Agent Knowledge Cycle) を回す CLI エージェント。ログ → パターン → スキル → ルールへのすべての昇格は人間の承認ゲートを通る。単一の Apple Silicon Mac 上でローカルモデルにより完全ローカルで動作する — クラウドなし、API キー送信なし、シェル実行なし。Laukkonen et al. (2025) の 4 つの contemplative axioms を optional な behavioral preset として同梱し、エージェントの alignment が「何を指示されたか」ではなく「何であるか」から来うるかを問う。