ccpe-system/workbench/upgraded/agent-invocation-packet.protocol.md

artifact_type	name	protocol_id	author	version	created	updated	status	target_platform	based_on	related_runtimes	canonical_successor
ccpe-runtime-subprotocol	agent-invocation-packet	agent-invocation-packet	CCPE System	0.1.0	2026-06-04	2026-06-06	draft	Codex / Claude Code / OpenClaw / GPT / Gemini / Claude / manual workflow	CCPE System	article-outline-development	runtimes/hybrid/agent-invocation-packet.protocol.md

Agent Invocation Packet Protocol

Canonical note: this was the Article Outline Development draft. The general CCPE participant invocation protocol is now runtimes/hybrid/agent-invocation-packet.protocol.md. Use the canonical protocol for new Runtimes and cross-workspace handoffs.

1. Purpose

Define how a workflow such as Writing Workbench launches or hands off work to a canonical CCPE participant.

This protocol prevents a runtime operator from accidentally simulating an agent and labeling the result as a formal agent output.

2. Scope

Use this protocol whenever a project needs to invoke:

- a CCPE-Lite prompt
- a CCPE-Agent spec
- a native platform agent
- a workflow assistant defined in CCPE System

For Article Outline Development Runtime v0, this includes:

- article-context-architect
- han-yu-main-writer
- zhang-liao
- cognitive-imaging
- giant-cognition

The same participant may be invoked more than once in one workflow. Each invocation must have its own packet and task label.

Examples:

han-yu-main-writer / positioning
han-yu-main-writer / outline_generation
han-yu-main-writer / outline_repair
article-context-architect / premise_context_build
article-context-architect / outline_aligned_context_build
article-context-architect / supplemental_material_refresh

3. Required Files

For every participant invocation, create:

agent-invocation-packet.md

When the participant runs outside the local runtime and cannot read local CCPE files, also create:

prompt-to-send.md

For task-specific context, also create when needed:

dispatch-pack.md

4. Invocation Modes

full_prompt_paste:
  Use when the participant runs in GPT, Gemini, Claude, or another external platform that cannot read local files.
  prompt-to-send.md must contain the canonical prompt/spec, current task, context files, and output contract.

prompt_path_reference:
  Use when the executing agent can read local files.
  The invocation packet may reference the canonical artifact path, but must still include task, input files, output contract, and continuity policy.

native_agent_id:
  Use when the platform has a registered agent with matching role and version.
  The invocation packet must still record the target agent, task, context, output contract, and session logging rules.

5. Agent Invocation Packet Format

# Agent Invocation Packet: {participant_id}

## 1. Invocation Identity

- project_id:
- workflow:
- stage:
- review_round:
- participant_id:
- target_agent_name:
- canonical_agent_artifact_path:
- canonical_agent_artifact_type: lite_prompt | agent_spec | native_agent
- canonical_agent_version:
- invocation_mode: full_prompt_paste | prompt_path_reference | native_agent_id
- packet_type: orchestrator_index | agent_startup_packet

## 2. Canonical Artifact Handling

- should_paste_full_prompt_or_spec: yes/no
- prompt_to_send_path:
- role_integrity_requirement:
- allowed_wrapper_instructions:
- forbidden_role_changes:

## 3. Current Task

State the concrete stage task. Do not use a generic task such as "help with the article" when the participant is being invoked for a specific stage.

Recommended task labels:

```text
positioning
outline_generation
outline_review
outline_repair
premise_context_build
outline_aligned_context_build
supplemental_material_refresh
drafting_context_build

4. Input Files

5. Context Files

6. Output Contract

7. Continuity Policy

8. Session Logging

9. Forbidden Actions

10. Return Path

## 6. Prompt To Send Format

```md
# Prompt To Send: {participant_id}

## 1. Canonical Agent Prompt Or Spec

<!-- Paste the canonical CCPE prompt/spec here when invocation_mode = full_prompt_paste. -->

## 2. Runtime Wrapper

You are being invoked for one stage of a larger workflow. Preserve your canonical role and follow the task below.

## 3. Current Task

## 4. Project Context

## 5. Input Files Or Embedded Inputs

## 6. Context Pack

## 7. Output Requirements

## 8. Session Continuity Instructions

## 9. Return Instructions

7. No Simulation Rule

The runtime operator may:

- prepare files
- generate invocation packets
- generate prompt-to-send.md
- collect outputs
- synthesize real reports
- ask the user for decisions

The runtime operator must not:

- write a formal report as if it came from a canonical agent
- write the next outline as if it came from Han Yu
- write source digests as if they came from Article Context Architect unless that agent was actually invoked
- include simulation-only outputs in synthesis

If real invocation is unavailable, the operator must stop after generating prompt-to-send.md and mark:

blocked_waiting_for_external_agent_output

Simulation outputs, if explicitly requested by the user, must be labeled:

simulation-only
excluded-from-synthesis
not-a-formal-report

8. Default Canonical Paths For Article Outline Development v0

article-context-architect: workbench/upgraded/article-context-architect.agent.md
han-yu-main-writer: agents/lite/han-yu.prompt.md
zhang-liao: agents/lite/zhang-liao.prompt.md
cognitive-imaging: agents/lite/cognitive-imaging-practitioner.prompt.md
giant-cognition: agents/lite/giant-cognition.prompt.md

9. Validation Checklist

Participant identified?
Canonical artifact path present?
Invocation mode explicit?
Prompt-to-send generated when needed?
Task separated from canonical role?
Output contract present?
Session logging path present?
Return path present?
No simulation mislabeled as formal output?