The agent description now states chain-of-thought is internal/not exposed, but this agent is also referenced by `reason.prompt.md` (which instructs the model to show reasoning at each step). This creates conflicting guidance about whether `agi-reasoning` outputs chain-of-thought, and can confuse both users and prompt selection. To resolve, either remove/soften the “not exposed” claim here, or introduce a separate agent for visible reasoning and update `reason.prompt.md` to use it.

[Bryan-Roe]

The agent description now states chain-of-thought is internal/not exposed, but this agent is also referenced by reason.prompt.md (which instructs the model to show reasoning at each step). This creates conflicting guidance about whether agi-reasoning outputs chain-of-thought, and can confuse both users and prompt selection. To resolve, either remove/soften the “not exposed” claim here, or introduce a separate agent for visible reasoning and update reason.prompt.md to use it.

Originally posted by @Copilot in #164 (comment)