Prompts

The prompts block toggles bundled “harness” prompt sections that get folded into the system prompt. Each field is optional and defaults to true; set one to false only when you explicitly want to omit that section.

Fields

Field	Default	Section
behaviour	true	High-level operating stance: inferring intent, staying scoped, surfacing risks, verifying work.
communication	true	How the agent talks to the user: concise, direct, no filler, clear final summaries.
execution	true	The understand → gather context → act → verify → report loop and tool-choice guidance.
guardrails	true	Security/risk posture: permission boundaries, security-request handling, prompt/data boundaries.

Coding and software-engineering guidance is not a prompt toggle — it lives in the built-in hooman-coding skill and loads automatically when relevant. Custom, always-on instructions belong in ~/.hooman/instructions.md, not this block.

Example configs

Defaults (all sections on — equivalent to omitting prompts entirely):

{
  "prompts": {
    "behaviour": true,
    "communication": true,
    "execution": true,
    "guardrails": true
  }
}

Drop the guardrails section (e.g. because your own instructions.md fully replaces it):

{
  "prompts": {
    "guardrails": false
  }
}

/config in an interactive chat session can toggle these without hand-editing JSON — see /config.