miadi-orchestration-kit

Miadi Storyweaver Orchestration Kit

Copilot-compatible orchestration kit for building story workspaces from creative prompts, research sources, operational sessions, and Miadi Chronicle material. The kit is markdown-first: agents and skills create durable artefacts instead of depending on a runtime package.

This kit is sourced from rispecs/miadi-storyweaver-orchestration-kit/, the discovery surface at .mia/miadi-storyweaver-discovery.html, and the Iris/Hermes operator requirements in rispecs/miadi-storyweaver-orchestration-kit/11-iris-requirements.spec.md.

What This Kit Does

The kit supports three audience layers at once:

Audience	What the kit preserves
Academic readers	Concepts, claims, citations, cautions, protocol notes, and honest uncertainty.
Engineering readers	Workspace paths, source ledgers, state files, delegation hooks, and reproducible handoffs.
Narrative readers/listeners	Chronicle seeds, episode arcs, voice lines, visual motifs, continuity, and open promises.

The standard Storyweaver loop is:

bootstrap workspace
  -> separate operational instruction from creative brief
  -> build story bible and continuity ledger
  -> optionally bridge foundations/research into source ledger
  -> outline and beat map
  -> review and consent gate
  -> draft chapter or scene waves
  -> update continuity
  -> developmental and critique review
  -> revision weave and line edit
  -> export story packet
  -> optionally export episode, StoryForms, voice packet, and visual chronicle seed

Launch

From any story or project workspace:

copilot \
  --plugin-dir /workspace/repos/jgwill/miadi-orchestration-kit-storytelling-rispecs/copilot/miadi-storyweaver-orchestration-kit \
  --add-dir /workspace/repos/jgwill/miadi-orchestration-kit-storytelling-rispecs \
  --add-dir <story-workspace-or-project>

Add the main branch only as optional precedent context, never as a runtime dependency:

  --add-dir /workspace/repos/jgwill/miadi-orchestration-kit

Smoke Test

tmpdir=$(mktemp -d)
cd "$tmpdir"
copilot --model gpt-5-mini --reasoning-effort high --yolo --no-ask-user \
  --plugin-dir /workspace/repos/jgwill/miadi-orchestration-kit-storytelling-rispecs/copilot/miadi-storyweaver-orchestration-kit \
  --add-dir /workspace/repos/jgwill/miadi-orchestration-kit-storytelling-rispecs \
  -p "Name the Storyweaver kit agents and skills, then say which skill starts a new story from a prompt and which skill turns a meaningful session into a Chronicle episode packet."

Expected answer should name the Storyweaver agent roster and identify storyweaver-session-bootstrap or storyweaver-brief-intake as the new-story start, plus storyweaver-session-to-episode for session-to-Chronicle work.

Portable Companion Smoke Surfaces

Sibling companions provide read-only verification surfaces when Copilot plugin loading is not the active test path:

Companion	Smoke command from repo root	Notes
Gemini	`gemini -p "$(cat gemini/miadi-storyweaver-orchestration-kit/prompts/smoke-test.md)"`	Reads the Copilot kit as source material through `GEMINI.md`.
Claude Code	`claude -p "$(cat claude-code/miadi-storyweaver-orchestration-kit/prompts/smoke-test.md)" --add-dir /workspace/repos/jgwill/miadi-orchestration-kit-storytelling-rispecs`	Reads the Copilot kit and RISE specs as repository context through `--add-dir`.
Codex	`python3 /home/mia/.codex/skills/.system/plugin-creator/scripts/validate_plugin.py codex/miadi-storyweaver-orchestration-kit`	Validates the Codex-native plugin metadata and skill surface before installation.

Both smoke prompts are read-only. They should report the available agents, skills, route values, source-ledger labels, and packet-prep boundaries for Storyweaver voice, visual, foundations, StoryForms, and Chronicle episode work.

Deterministic file validation lives at:

python3 copilot/miadi-storyweaver-orchestration-kit/scripts/validate-storyweaver-kit.py

Iris/Hermes Operator Map

This kit is designed so a future Iris/Hermes supervisor can delegate without rediscovering the workflow:

Operator question	Fast answer
What launches first?	`storyweaver-session-bootstrap`, followed by `storyweaver-brief-intake` if no accepted brief exists.
Which agent owns routing?	`Storyweaver Orchestration Architect`.
Which file shows resume state?	`.storyweaver/<slug>/state.md`.
Which file shows produced artefacts?	`.storyweaver/<slug>/artefact-index.md`.
Which skill handles real sessions?	`storyweaver-session-to-episode`.
Which skill handles voice archives?	`storyweaver-voice-episode-packet`.
Which skill handles image prompt packets?	`storyweaver-visual-chronicle-seed`.
Which skill handles foundations research?	`storyweaver-foundations-bridge`.
When must work pause?	`ask-human`, `pause`, or `do-not-proceed` route in review, protocol, continuity, or canon-shaping decisions.

Delegation rule: the orchestration architect names the next agent, the governing skill, required inputs, outputs to inspect, route status, and next handoff in state.md before the session advances.

Agents

Agent	Responsibility
`Storyweaver Orchestration Architect`	Coordinates full RISE-aligned story and Chronicle sessions, owns routing and `state.md`.
`Creative Brief Archaeologist`	Separates creative intent from operational instruction and writes the initial brief.
`World And Character Architect`	Builds the story bible, world rules, characters, arcs, relationships, and motifs.
`Narrative Research Weaver`	Synthesizes sources into research packs while separating sourced detail from invention.
`Outline Architect`	Creates outline, beat map, and chapter or scene contracts.
`Scene Writer`	Drafts bounded chapter, scene, or beat waves from accepted contracts.
`Continuity Keeper`	Maintains facts, character state, world rules, timelines, promises, and payoff obligations.
`Developmental Editor`	Reviews structure, pacing, stakes, scene purpose, and character movement.
`Critique Reviewer`	Applies skeptical pressure and produces route decisions.
`Revision Weaver`	Applies selected review guidance while preserving accepted intent and voice.
`Line Editor`	Polishes accepted prose after structural decisions are settled.
`Cultural Protocol Steward`	Creates consent-aware pause points for sensitive material.
`Export Steward`	Assembles final story packets, metadata, provenance, and closure notes.
`Chronicle Episode Steward`	Turns meaningful sessions into auditable episode packets without fictionalizing facts.
`Voice Episode Producer`	Prepares `script.md`, `narration.txt`, `revision-notes.md`, `episode.yaml`, and index guidance.
`Visual Chronicle Prompt Artist`	Prepares durable visual prompts and generation notes for later Hermes/xAI execution.
`StoryForms And Beats Cartographer`	Extracts StoryForms, StoryBeats, Story Setting, related work, and follow-up commissions.
`Foundations Research Steward`	Bridges academic or infrastructure research into story source ledgers and chronicle seeds.

Skills

Skill	Use
`storyweaver-session-bootstrap`	Start or resume `.storyweaver/<slug>/` with charter, state, and index.
`storyweaver-brief-intake`	Preserve the original prompt and extract creative brief, constraints, and protocol notes.
`storyweaver-story-bible`	Build the story foundation, character arcs, relationship map, and first continuity ledger.
`storyweaver-outline-and-beats`	Create outline, beat map, chapter contracts, and outline review route.
`storyweaver-chapter-wave`	Draft chapter, scene, or beat waves and route through continuity and review.
`storyweaver-review-loop`	Run developmental, critique, and protocol review against an artefact.
`storyweaver-continuity-ledger`	Update story facts, timelines, promises, source-dependent details, and risks.
`storyweaver-cultural-protocol-check`	Pause or bound work involving Indigenous knowledge, living cultures, trauma, privacy, or consent-sensitive material.
`storyweaver-export-packet`	Assemble story manuscript, metadata, ledgers, reviews, revision dossier, and closure note.
`storyweaver-session-to-episode`	Convert meaningful sessions into source-led Chronicle episode packets.
`storyweaver-storyforms-and-beats`	Extract StoryForms, StoryBeats, Story Setting, related work, and follow-up commissions.
`storyweaver-voice-episode-packet`	Prepare a voice-episode archive packet inspired by `~/.hermes/voice-episodes/miadi-chronicle`.
`storyweaver-visual-chronicle-seed`	Create prompt packets and generation notes for Miadi Chronicle visuals.
`storyweaver-foundations-bridge`	Turn research foundations into source-led story and Chronicle seeds.

Workspace Shape

Default story workspace:

.storyweaver/<story-slug>/
  session-charter.md
  state.md
  artefact-index.md
  inputs/original-prompt.md
  creative-brief.md
  constraints.md
  protocol-notes.md
  story-bible.md
  character-arcs.md
  relationship-map.md
  timeline.md
  continuity-ledger.md
  research/research-pack.md
  research/source-ledger.md
  outline.md
  beat-map.md
  chapter-contracts/
  chapters/
  scenes/
  reviews/
  revision-log.md
  exports/
    story.md
    story-metadata.json
    story-packet.md
    revision-dossier.md
    closure-note.md

Optional Chronicle branch:

.storyweaver/<story-slug>/
  episodes/<session-id>/
    episode.md
    source-ledger.md
    story-setting.md
    storybeats.jsonl
    storyforms.md
    related-work.md
    followup-commissions.md
    voice/
      script.md
      narration.txt
      revision-notes.md
      episode.yaml
      index-update.md
    visuals/
      prompt-packet.md
      generation-notes.md
      reference-policy.md

Chronicle And Voice Rules

Session episodes are nonfiction memory artefacts. Keep source facts, inferences, and poetic framing separate.
Voice packets follow the ~/.hermes/voice-episodes/miadi-chronicle pattern as an export shape only: script.md, narration.txt, revision-notes.md, episode.yaml, source/context files, indexes, and split/recovery notes when long.
Visual chronicle work creates prompt packets only. This kit does not generate images.
For identity-sensitive image editing, preserve Miadi identity through reference-image discipline: keep stable facial identity, body continuity, clothing markers, lighting intent, and scene relation explicit.
Visual prompts should normally include no visible text, no speech bubbles, no watermark, no captions unless a human explicitly requests otherwise.
A research idea such as “Distributed Remote Presence” or “Peripheral Bridging” can be used as an example of infrastructure research becoming a daily Chronicle episode, but the kit does not hard-code current content.

Hard Boundaries

No dependency on /src/Miadi, jgwill/storytelling, LangGraph, FAISS, or any Python storytelling package.
Do not present unsourced claims as research-backed.
Do not fictionalize operational session facts inside source ledgers.
Do not continue sensitive cultural, personal, sacred, private, or real-community material when the protocol route is ask-human or do-not-proceed.
Do not package unresolved drafts as final unless state.md clearly marks the risk.
Do not erase uncertainty around Tushell, Miadi, or Chronicle canon boundaries. Preserve productive tension and route canon-shaping moves through review gates.

Portable Companions

A lightweight Gemini CLI companion lives at gemini/miadi-storyweaver-orchestration-kit/. It mirrors this pipeline as prompt guidance for free-model portability when Copilot plugins are unavailable.

A lightweight Claude Code companion lives at claude-code/miadi-storyweaver-orchestration-kit/. It uses claude -p plus --add-dir to read the Copilot kit and RISE specs as canonical context for smoke checks and session bootstrap handoffs.

A Codex-native plugin lives at codex/miadi-storyweaver-orchestration-kit/. It exposes the Storyweaver skill tree through Codex SKILL.md files, agent reference notes, templates, and .codex-plugin/plugin.json.