`reason.prompt.md` is described here as showing explicit chain-of-thought steps to the user, but `reason.prompt.md` currently uses `agent: agi-reasoning` and that agent’s front-matter description now says chain-of-thought is not exposed. This is internally inconsistent and may mislead users/tool routing. Consider either (a) adjusting this prompt list entry to match the agent’s behavior, or (b) splitting into two agents (hidden-CoT vs visible-reasoning) and pointing `reason.prompt.md` at the visible one.

[Bryan-Roe]

reason.prompt.md is described here as showing explicit chain-of-thought steps to the user, but reason.prompt.md currently uses agent: agi-reasoning and that agent’s front-matter description now says chain-of-thought is not exposed. This is internally inconsistent and may mislead users/tool routing. Consider either (a) adjusting this prompt list entry to match the agent’s behavior, or (b) splitting into two agents (hidden-CoT vs visible-reasoning) and pointing reason.prompt.md at the visible one.

Originally posted by @Copilot in #164 (comment)