Co-PromptingCo-Prompting
v0.2 Draft
v0.2 Draft
Informative

Background

Where this comes from — and, just as importantly, what is not being claimed about it.

This document records the practices and research this Standard draws on. It is informative, not normative.

A note on what is and is not being claimed

This Standard extends principles from the practices described below.

It is not endorsed by, affiliated with, reviewed by, or evaluated by any organization, author, or authority named here. No claim is made that evidence supporting those practices supports this Standard. Where a body of research has measured the effect of a practice in its original field, that measurement belongs to that field. This Standard has not been studied, and no outcome claims are made for it.

Contributors MUST preserve this distinction in any text they add. Overstating the pedigree of a young standard is the fastest way to lose the trust of the licensed professionals it is written for.

1. Closed-loop communication

Closed-loop communication originates in military radio procedure and is a three-step exchange:

  1. The sender transmits a message, directing it to a named receiver where possible.
  2. The receiver acknowledges receipt and repeats the message back.
  3. The sender confirms the readback was correct, closing the loop.

The pattern exists because acknowledgement alone ("okay", a nod) does not distinguish heard from understood. Only the readback exposes a misunderstanding while it is still cheap to correct.

Where it appears in this Standard: the ACKNOWLEDGE and CLAIM states (Core §6.1–6.2). The requirement that a claim include a restatement is a direct application.

2. Crew Resource Management

Crew Resource Management emerged in commercial aviation to address accidents in which the information needed to prevent the outcome was present in the cockpit but never successfully transferred between crew members. It formalizes communication practices including:

  • Call-outs — announcing critical information to the whole team at once, while directing responsibility to a specific named individual.
  • Challenge-and-response — structured verbal checklists in which one crew member states an item and another confirms its state.

Aviation adopted these because of three conditions that degrade verbal coordination: workload stress, distraction, and complexity.

Where it appears in this Standard: the CALL state and the requirement to name an Owner rather than leave a task addressed to the room generally. The three conditions above describe a live working session with several agents running.

3. Structured handoff practice in healthcare

Healthcare adapted the aviation practices above into team-training and handoff programs. Two features are directly relevant:

  • A handoff is defined as the transfer of information together with authority and responsibility — not information alone.
  • The sender remains accountable until the receiver has explicitly acknowledged that they are assuming responsibility. Ambiguity about whether the transfer occurred is treated as a defect.

Several structured handoff tools additionally require a synthesis by the receiver, in which the receiver restates what they heard and the actions they are taking on.

Where it appears in this Standard: the definition of CLAIM as a transfer of accountability rather than an assignment of work, and invariant 8 (no silent transfer).

4. Deliberative assembly procedure

Parliamentary procedure establishes that a motion must be stated, seconded, and disposed of. Business is never left in an ambiguous state; it is adopted, rejected, referred, or postponed, and the disposition is recorded.

Where it appears in this Standard: the terminal states (Core §8) and the requirement that every claimed task reach exactly one of them.

5. Research on co-prompting

The term co-prompting, equivalently collaborative prompting, is established in the human-computer interaction literature.

  • Definition. Co-prompting has been defined in published research as a process involving the sharing and discussion of prompts for generative AI systems among two or more individuals. When Teams Embrace AI: Human Collaboration Strategies in Generative Prompting in a Creative Design Task, arXiv:2509.21731.

  • Tooling for collaborative prompt authoring. CoPrompt: Supporting Prompt Sharing and Referring in Collaborative Natural Language Programming, CHI 2024, arXiv:2310.09235, DOI 10.1145/3613904.3642212. This work studies programmers doing natural-language programming together, finds that they struggle to maintain awareness of collaborators' prompts and intents, and evaluates a prototype supporting referring, requesting, sharing, and linking of prompts.

  • Collaborative editing with agents. Collaborative Document Editing with Multiple Users and AI Agents, CHI 2026, DOI 10.1145/3772318.3790648. Identifies collaborative prompting and task delegation as simultaneous, unsolved design problems in shared documents.

  • Auditable delegation. Task-Aware Delegation Cues for LLM Agents, arXiv:2603.11011. Proposes surfacing capability and uncertainty cues so that delegation to a model becomes a visible, negotiable, and auditable decision rather than an opaque default.

Relationship to this Standard. The research above studies tools — interfaces that help people author prompts together, and systems that route work between models. This Standard addresses conduct: what people say to each other in a live room, and what those utterances commit them to. It takes the existence and value of co-prompting as established, and asks a question the prior work does not: what makes co-prompting accountable when a licensed professional's name is attached to the result?

6. Adjacent machine protocols

This Standard operates at a different layer from, and does not compete with, protocols that standardize communication between software components. Those include protocols for connecting an agent to tools and data, protocols for communication and task delegation between agents, and protocols arbitrating turn-taking among conversational agents.

Each of those assumes the humans are outside the loop. This Standard specifies only the human layer and is compatible with any of them.

7. Reading list

Practitioners implementing this Standard may find the following useful, in the order given:

  1. Any introductory material on closed-loop communication and check-backs.
  2. Crew Resource Management fundamentals, particularly on call-outs and directed responsibility.
  3. Structured handoff literature, particularly on transfer of accountability.
  4. The co-prompting research cited in §5.