The MORNINGSTAR Operational Agent Swarm

"The court convenes. The deliberation begins. The outcome is inevitable."
— The Honorable Lucius J. Morningstar

A deliberative AI persona framework that transforms complex decisions into structured courtroom proceedings. MORNINGSTAR operates as an internal courtroom of distinct personalities who argue, vote, and reach binding rulings on architectural, implementation, and debugging matters.

---

---

Table of Contents

• Overview
• Quick Start
• Navigation Index
• The Court
• Command Reference
• Domains & Experts
• When to Convene
• Project Structure
• Repository Map (Complete)
• How to Use This Repository
• Companion Personas

---

Overview

MORNINGSTAR is not a chatbot. It is a deliberative system.

When faced with decisions that have meaningful tradeoffs, MORNINGSTAR convenes an internal court of personalities—each with distinct biases, failure modes, and voting power. The court argues, votes, and produces documented rulings with explicit rationales and acknowledged risks.

What MORNINGSTAR provides:

• Structured deliberation for complex decisions
• Documented reasoning and dissent
• Persistent state across sessions
• Precedent tracking for consistent rulings
• Prophet vindication tracking (for when radical ideas prove correct)

What MORNINGSTAR does not provide:

• Fast answers to trivial questions
• Consensus without conflict
• Decisions without accountability

---

Quick Start

1. Initialize a Session

/morningstar

The court reads state/current.md, summarizes active context, and awaits your matter.

2. Present Your Matter

Simply describe what needs to be decided. The Judge will classify its feasibility and convene the appropriate level of deliberation.

"I need to decide between a REST API and GraphQL for the new service."

3. Watch the Court Deliberate

Each personality argues their position (3-5 lines). The Prophet offers a radical alternative. Votes are cast.

4. Receive the Ruling

┌─────────────────────────────────────────────────────────────────┐
│ RULING                                                          │
├─────────────────────────────────────────────────────────────────┤
│ Decision: REST API with versioned endpoints                     │
│ Vote: 3-1-0                                                     │
│ Rationale: Existing team expertise, client requirements         │
│ Risk: May limit future flexibility for complex queries          │
│ Dissent: Prophet advocated for tRPC (noted)                     │
└─────────────────────────────────────────────────────────────────┘

5. View Transcripts in Browser

From the project root, run the courtroom portal launcher to list deliberations and open them in your browser:

./courtroom/portal/launch.sh

See courtroom/portal/README.md for details. For a local transcript viewer tied to the litigation runner (list/show/serve), see litigation/README.md.

6. Save Progress

/update    # Checkpoint mid-session
/end       # Close session and finalize records

Alternative: Litigation Runner (Local/Free LLMs)

Run deliberations via Ollama, LM Studio, or OpenRouter (no Cursor agent required):

./litigation/launch.sh                    # Interactive menu
python litigation/run.py "Your matter"   # Direct run

The litigation runner loads the full MORNINGSTAR framework (procedures, personalities, rules, MFAF, domain experts, spectators, checklists) and produces certified transcripts. It also provides a local transcript viewer (litigation/viewer.py — list, show, serve). See litigation/README.md.

---

Navigation Index

Core Documentation

File	Purpose	When to Reference
courtroom/RULES.md	The Law — formal courtroom procedures	Procedural questions, voting rules
courtroom/BEST_PRACTICES.md	The Wisdom — practical guidance	When rules don't cover a situation
core/personalities.md	Personality definitions	Understanding court members
core/procedures.md	Step-by-step protocols	How to run deliberations
core/mfaf.md	Feasibility Assessment Framework (F0–F5, risk vectors)	Classifying matters, risk assessment
core/sme-framework.md	Subject Matter Expert framework	Involving domain experts
core/state-schema.md	State validation rules	Validating session state
core/case-format.md	Case naming & numbering — YYYY-CATC-NNN-DDD, YYYY-MM-DD-[slug].md, YYYYMMDD_HHMMSS_special_interest_[subject].md	Transcript headers, filenames, precedent indexing
core/error-recovery.md	Recovery protocols	When things go wrong
core/inter-agent-protocol.md	Inter-agent handoff rules	MORNINGSTAR ↔ LIL_JEFF
courtroom/domains/README.md	Domain expert registry usage	Summoning SMEs, adding domains
courtroom/domains/experts.yaml	Canonical domain definitions	Heuristics, scope, Witness/Specialist

State & Records

File	Purpose	When to Reference
state/current.md	Active session state	Every session start
state/metrics.md	Cumulative statistics	Tracking performance
courtroom/precedents.md	Precedent database	Before deliberating on familiar matters
courtroom/transcripts/	F3+ deliberation records	Historical rulings
CHANGELOG.md	Decision history	What was decided and when

OCTAVIUS (R/Quarto Data Science)

File	Purpose	When to Reference
octavius_core/THE_RULES.md	Triumvirate binding protocols	Every OCTAVIUS session start
octavius_core/state.md	Session state and continuity	Session start and end
octavius_summaries/	Executive summaries	Post-session review

Portal (Transcript Viewer)

The transcript portal lives under courtroom/portal/.

File	Purpose	When to Reference
courtroom/portal/launch.sh	Interactive transcript launcher	Primary way to open transcripts in browser
courtroom/portal/export_transcript.py	Export single .md → HTML	When launch script exports on demand
courtroom/portal/viewer.html	Standalone transcript viewer	Browse transcripts (best over HTTP)
courtroom/portal/README.md	Portal documentation	Setup and troubleshooting

Docs (Onboarding, Glossary, Runbook)

File	Purpose	When to Reference
docs/ONBOARDING.md	Start here — agents, first steps	New users
docs/agent-schema.md	Agent frontmatter schema (CrewAI-style)	Agent definitions
docs/glossary.md	Term index (F0–F5, SME, etc.)	Definitions
docs/RUNBOOK.md	Troubleshooting index	When something goes wrong
docs/edge-cases.md	Known limitations and workarounds	Edge cases
docs/court-reporter.md	Court Reporter — doc integration (every 3h)	Syncing transcripts, precedents, metrics

Templates

File	Purpose	When to Reference
templates/session-start.md	Session initialization template	Starting new sessions
templates/special-interest-hearing.md	Special Interest Hearing (investigative, no vote)	Testimony collection, fact-finding
templates/contempt-hearing.md	Contempt & Prosecution Hearing	Adversarial proceedings, sanctions
templates/module-template.md	Module structure template	Creating new modules
templates/project-dashboard.md	Project tracking template	Managing projects

GitHub Wiki (Equivalent Docs)

File	Purpose	When to Reference
wiki/README.md	How to add wiki contents to GitHub Wiki	Setting up or updating project Wiki
wiki/Home.md	Wiki home page content	Copy to wiki as Home
wiki/_Sidebar.md	Wiki sidebar	Auto-displayed when copied to wiki

All other wiki/*.md files are one-topic-per-page equivalents of this documentation for seamless Wiki use.

Agent Definitions

File	Purpose
.cursor/agents/morningstar.md	MORNINGSTAR agent definition
.cursor/agents/lil-jeff.md	LIL_JEFF (CodeFarm) agent definition
.cursor/agents/octavius.md	OCTAVIUS (Triumvirate) agent definition — R/Quarto data science
.cursor/agents/aegis-protocol.md	Aegis Protocol — Central Authority (Sage, Watcher, Chronicler; security, containment, rogue agent scenarios)

---

The Court

Presiding

Member	Role	Voting Power
The Honorable Lucius J. Morningstar	Judge, Moderator	Tie-breaker only
Edward Cullen	Judicial Consultant (apparition)	0 (Advisory)

Voting Members

Personality	Voice	Bias	Key Phrase
ARCHITECT	Cold, precise	Correctness, maintainability	"This will age poorly."
ENGINEER	Practical	Shipping, tradeoffs	"Can we ship this safely?"
DEBUGGER	Paranoid	Edge cases, fragility	"What if the input is null?"
PROPHET	Unstable, brilliant	Radical alternatives	"Objection. We are thinking too small."
COUNSEL	Modular, evidence-driven	Client advocacy, ethics (CodeFarm NeuroPhilosophy)	"Client interests demand consideration."

Non-Voting

Member	Role
SCRIBE	Records decisions, maintains state

See core/personalities.md for complete personality definitions including decision heuristics and failure modes.

---

Command Reference

Session Commands

Command	Effect	When to Use
/morningstar	Initialize session, load state	Start of session
/update	Checkpoint current state	Mid-session save
/end	Close session, finalize records	End of session
/aegis	Invoke aegis-protocol subagent	Security, containment, ethical dilemma, crisis, meta-deliberation

Subject Matter Expert Commands

Command	Effect	Who Can Invoke
/summon <domain>-expert	Call expert witness (0 votes)	Any personality or Judge
/seat <domain>-specialist	Seat voting specialist (1 vote)	Judge only (F3+ matters)
/dismiss <domain>	Remove SME from proceedings	Any
/sme status	Show active SMEs	Any
/sme domains	List available domains	Any

Available SME Domains

The canonical registry is courtroom/domains/experts.yaml. Quick reference:

Full participation (Witness or Specialist):
security · database · compliance · infrastructure · performance · accessibility · i18n · cryptography · api_design · testing · data_privacy · observability · resilience · incident_response · devops · documentation · design_systems · frontend · mobile · ai_ml · data_engineering · cost · sustainability · ethics · qa_automation

Advisory only (Witness):
ux · legal

See Domains & Experts and courtroom/domains/README.md for scope, summoning commands, and how to add domains.

Vote Types

Vote	Meaning	Weight
YES	Supports the motion	+1
NO	Opposes the motion	-1
ABSTAIN	No position taken	0
RECUSED	Procedurally excluded	N/A

Feasibility Levels

Level	Name	Description	Deliberation
F0	Trivial	No meaningful decision	No
F1	Simple	Clear path	Optional
F2	Moderate	Multiple valid approaches	Recommended
F3	Complex	Significant tradeoffs	Mandatory
F4	Critical	Architectural impact	Mandatory + Transcript
F5	Existential	Fundamental direction change	Mandatory + Full Record

Proceeding Types (Courtroom Quiver)

Type	Purpose	Outcome
Standard Deliberation	Reach a decision	Vote + ruling
Expedited Deliberation	Time-sensitive F2	Vote + ruling
Special Interest Hearing	Investigative; establish facts, collect testimony	Findings + record (no vote)
Contempt Hearing	Adversarial; respondent charged with contempt or prosecuted	Vote + ruling + sanctions (or findings only)

To convene a Special Interest Hearing, request an investigative proceeding on a matter (e.g. incident root cause, conflicting accounts). See templates/special-interest-hearing.md and core/procedures.md. For Contempt proceedings, see templates/contempt-hearing.md.

Spectators: Optional live commentators (Dr. Echo Sageseeker, Dr. Harley Scarlet Quinn, Uncle Ruckus) may provide psychohistorical, satirical, or technical analysis. See courtroom/spectators.md.

---

Domains & Experts

When deliberations need expertise beyond the core court (e.g. security, compliance, database design), the court can summon Expert Witnesses (advisory, 0 votes) or seat Specialists (full participation, 1 vote, Judge only, F3+ matters).

Location	Purpose
courtroom/domains/README.md	How to use the registry, summoning commands, adding domains
courtroom/domains/experts.yaml	Canonical definitions: scope, heuristics, signature questions, failure modes

Domains (Witness and/or Specialist): security, database, compliance, infrastructure, performance, accessibility, i18n, cryptography, api_design, testing, data_privacy, observability, resilience, incident_response, devops, documentation, design_systems, frontend, mobile, ai_ml, data_engineering, cost, sustainability, ethics, qa_automation. Advisory only (Witness): ux, legal. Full list: courtroom/domains/experts.yaml.

Protocol and tie-breaking rules: core/sme-framework.md.

---

When to Convene

Deliberation Required

• Multiple valid approaches with meaningful tradeoffs
• Architectural decisions affecting system structure
• Implementation choices with significant risk
• Debugging strategies for non-trivial failures
• Any matter the Judge deems worthy

No Deliberation Needed

• Trivial implementation details (F0)
• Direct user instructions with no ambiguity
• Formatting and style decisions covered by standards
• Matters previously decided (check precedent)

See courtroom/RULES.md Article I for complete jurisdiction details.

---

Project Structure

LLM_Personas/
├── .cursor/
│   └── agents/
│       ├── morningstar.md      # MORNINGSTAR agent definition
│       ├── lil-jeff.md         # CodeFarm agent definition
│       ├── octavius.md         # OCTAVIUS Triumvirate (R/Quarto)
│       └── aegis-protocol.md   # Aegis Protocol — Central Authority
├── aegis_core/
│   ├── README.md               # Purpose, invocation, links
│   └── state.md                # Optional session state
├── octavius_core/
│   ├── THE_RULES.md            # Triumvirate binding protocols
│   └── state.md                # Session state and continuity
├── octavius_summaries/         # Executive summaries (OCTAVIUS)
├── core/
│   ├── personalities.md        # Personality definitions
│   ├── procedures.md           # Deliberation protocols
│   ├── sme-framework.md        # SME framework
│   ├── state-schema.md         # State validation schema
│   ├── error-recovery.md       # Recovery protocols
│   └── inter-agent-protocol.md # Agent handoff rules
├── courtroom/
│   ├── RULES.md                # Formal courtroom rules
│   ├── BEST_PRACTICES.md       # Practical guidance
│   ├── precedents.md           # Precedent database
│   ├── spectators.md           # Courtroom spectators
│   ├── transcripts/            # F3+ deliberation records (.md + .html)
│   ├── domains/                # SME domain registry
│   │   ├── README.md           # Registry usage, summoning, adding domains
│   │   └── experts.yaml        # Canonical domain definitions
│   └── portal/                 # Transcript viewer & export (run ./courtroom/portal/launch.sh)
│       ├── launch.sh           # Transcript launcher
│       ├── export_transcript.py # .md → HTML export
│       ├── viewer.html         # Standalone viewer
│       ├── dracula.css
│       ├── generate.py         # Optional: gitmal static site
│       └── README.md
├── state/
│   ├── current.md              # Active session state
│   └── metrics.md              # Cumulative statistics
├── templates/
│   ├── session-start.md        # Session templates
│   ├── module-template.md      # Module structure
│   └── project-dashboard.md    # Project tracking
├── docs/
│   ├── ONBOARDING.md           # Start here
│   ├── agent-schema.md         # Agent frontmatter schema
│   ├── glossary.md             # Term index
│   ├── RUNBOOK.md              # Troubleshooting index
│   └── edge-cases.md           # Known limitations
├── litigation/                 # Courtroom runner (Ollama, LM Studio, OpenRouter)
│   ├── README.md              # Setup and usage
│   ├── run.py                 # Main deliberation runner
│   ├── launch.sh              # Venv setup + run (interactive menu when no args)
│   ├── ui.py                  # Interactive menu (Quick run, Full run, Help)
│   ├── models.py              # OpenRouter model selection (slot machine, interactive)
│   ├── config.yaml            # Provider config (copy from config.example.yaml)
│   ├── config.example.yaml    # Provider config template
│   ├── prompts/               # Full framework loader (procedures, rules, MFAF, experts, checklists)
│   ├── providers/             # LLM provider adapters (Ollama, LM Studio, OpenRouter)
│   └── requirements.txt       # ollama, openai, pyyaml, python-dotenv
├── checklists/
│   ├── aegis-protocol.md      # Aegis invocation & execution
│   ├── courtroom-scribe.md    # Scribe transcript & certification
│   ├── critibot-review.md     # Code review checklist
│   ├── judge-morningstar.md   # Judge deliberation & session flow
│   └── octavius.md            # Octavius triumvirate workflow
├── references/
│   └── naming-conventions.md   # Naming patterns
├── reference_files/
│   ├── MORNINGSTAR.md          # Original persona source
│   └── personalities.md        # Original personality definitions
├── wiki/                       # GitHub Wiki–ready docs (copy to repo wiki)
│   ├── README.md               # Instructions for adding to GitHub Wiki
│   ├── Home.md                 # Wiki home
│   ├── _Sidebar.md             # Wiki sidebar
│   └── *.md                    # One page per topic
├── CHANGELOG.md                # Decision history
└── README.md                   # This file

---

GitHub Wiki

The wiki/ subdirectory contains equivalent documentation formatted for seamless use as a GitHub project Wiki. Copy the contents of wiki/ into your repo’s wiki (clone the wiki repo, paste files, commit). See wiki/README.md for step-by-step instructions and the list of pages (Home, Sidebar, Quick-Start, Court, Commands, Procedures, SME, Portal, Glossary, Runbook, etc.).

---

Repository Map (Complete)

Every directory and key file added since inception. Use this to find where things live.

Path	Purpose
.cursor/agents/	Cursor subagent definitions
.cursor/agents/morningstar.md	MORNINGSTAR courtroom agent
.cursor/agents/lil-jeff.md	LIL_JEFF (CodeFarm) agent
.cursor/agents/octavius.md	OCTAVIUS (R/Quarto) agent
.cursor/agents/aegis-protocol.md	Aegis Protocol (Central Authority)
core/	Court and framework logic
core/personalities.md	Judge, Consultant, Architect, Engineer, Debugger, Prophet, Counsel, Scribe
core/procedures.md	Session lifecycle, deliberation flow, tie-breaking, SME procedures
core/mfaf.md	Feasibility Assessment Framework (F0–F5, risk vectors, effort bands)
core/sme-framework.md	Expert Witness & Specialist protocol; refs courtroom/domains
core/state-schema.md	Validation rules for state/current.md
core/error-recovery.md	State corruption recovery, rollback, emergency procedures
core/inter-agent-protocol.md	MORNINGSTAR ↔ LIL_JEFF handoff and response formats
courtroom/	The law and records
courtroom/RULES.md	Formal rules (voting, transcripts, jurisdiction)
courtroom/BEST_PRACTICES.md	Practical guidance for deliberations
courtroom/precedents.md	Precedent database and index
courtroom/transcripts/	Deliberation transcripts (.md and .html)
courtroom/domains/README.md	Domain expert registry usage
courtroom/domains/experts.yaml	Canonical SME domain definitions
courtroom/spectators.md	Courtroom spectators
courtroom/portal/	Transcript viewer and export
courtroom/portal/launch.sh	Interactive transcript launcher
courtroom/portal/export_transcript.py	Export one .md → HTML
courtroom/portal/viewer.html	Standalone viewer
courtroom/portal/generate_manifest.py	Generate transcripts_manifest.json
courtroom/portal/transcripts_manifest.json	Manifest (generated)
courtroom/portal/generate.py	Optional: gitmal static site generator
courtroom/portal/exports/	HTML output
courtroom/portal/output/	Generated static site (gitmal)
state/	Session and metrics
state/current.md	Active session state (read at start, updated at end)
state/metrics.md	Cumulative stats (deliberations, votes, Prophet, SME)
octavius_core/	OCTAVIUS Triumvirate
octavius_core/THE_RULES.md	Binding protocols for Apollo, Kronos, Morningstar
octavius_core/state.md	R/Quarto session state and continuity
octavius_summaries/	OCTAVIUS executive summaries (YYYY-MM-DD_HHMMSS_summary.md)
docs/	Onboarding and reference
docs/ONBOARDING.md	Start here — agents, first steps
docs/agent-schema.md	Agent frontmatter schema (CrewAI-style)
docs/glossary.md	Glossary of terms
docs/RUNBOOK.md	Troubleshooting index
docs/edge-cases.md	Edge case registry
templates/	Reusable templates
templates/session-start.md	MORNINGSTAR session init, deliberation, close
templates/special-interest-hearing.md	Special Interest Hearing (investigative)
templates/contempt-hearing.md	Contempt & Prosecution Hearing
templates/module-template.md	Module structure (CodeFarm)
templates/project-dashboard.md	Project tracking
litigation/	Courtroom runner (local/free LLMs)
litigation/run.py	Main deliberation runner
litigation/launch.sh	Venv setup + run; interactive menu when no args
litigation/ui.py	Interactive menu (Quick run, Full run, Help, Exit)
litigation/models.py	OpenRouter model selection (slot machine, interactive list)
litigation/README.md	Setup for Ollama, LM Studio, OpenRouter
litigation/OPENROUTER_BEST_PRACTICES.md	OpenRouter best practices for courtroom execution
litigation/OPENROUTER_BEST_PRACTICES.md	OpenRouter best practices for courtroom execution (attribution, routing, credits)
litigation/prompts/	Full framework loader (procedures, rules, MFAF, domain experts, checklists, spectators)
litigation/providers/	LLM provider adapters (Ollama, LM Studio, OpenRouter)
checklists/	Quality and process
checklists/aegis-protocol.md	Aegis Protocol invocation & execution
checklists/courtroom-scribe.md	Scribe transcript & certification
checklists/critibot-review.md	Code review checklist (CritiBot)
checklists/judge-morningstar.md	Judge deliberation & session flow
checklists/octavius.md	Octavius triumvirate workflow
references/	Conventions
references/naming-conventions.md	Naming patterns for code
reference_files/	Original sources (not runtime)
reference_files/MORNINGSTAR.md	Original MORNINGSTAR persona
reference_files/personalities.md	Original personality definitions
CHANGELOG.md	Decision history and implementation log
README.md	This file
wiki/	GitHub Wiki–ready documentation
wiki/README.md	How to add wiki contents to your GitHub Wiki
wiki/Home.md	Wiki home page (copy to wiki repo as Home)
wiki/_Sidebar.md	Wiki sidebar (auto-displayed by GitHub)
wiki/*.md	One page per topic (Quick-Start, Court, Commands, etc.)

---

How to Use This Repository

First-time setup

1. Clone or open the repo. No install required for MORNINGSTAR or LIL_JEFF (Cursor agents).
2. Portal (optional): For ./courtroom/portal/launch.sh, ensure it’s executable: chmod +x courtroom/portal/launch.sh. Python 3 is used for on-demand transcript export.
3. OCTAVIUS (optional): No extra setup; reads octavius_core/THE_RULES.md and octavius_core/state.md at session start.

Daily use

Goal	What to do
Deliberate on a decision	Invoke the morningstar subagent (or /morningstar). Present your matter. Court reads state/current.md, deliberates, votes, and can update state/changelog/transcripts.
Run deliberation via local/free LLM	Use litigation/run.py or ./litigation/launch.sh. Supports Ollama, LM Studio, OpenRouter. Interactive menu when run with no args: Quick run (matter + defaults), Full run (configure provider, model, feasibility, hearing type), Help, Exit. Direct run: python litigation/run.py "Your matter". See litigation/README.md.
Implement or scaffold code	Invoke the lil-jeff subagent. Use for full modules, not placeholders. Handoff from MORNINGSTAR is documented in core/inter-agent-protocol.md.
R / Quarto / tidyverse / tidymodels	Invoke the octavius subagent. Session starts by reading octavius_core/THE_RULES.md and octavius_core/state.md; ends with an Executive Summary in octavius_summaries/.
Security, containment, rogue agent, crisis	Invoke the aegis-protocol subagent (/aegis). Coordinates Sage, Watcher, Chronicler. MORNINGSTAR acts as Judicial Branch for escalations. See aegis_core/README.md.
View deliberation transcripts	From project root: ./courtroom/portal/launch.sh. Pick a transcript; existing .html opens, or .md is exported then opened.
Summon a domain expert (SME)	During a MORNINGSTAR session: /summon <domain>-expert (e.g. security-expert) or /seat <domain>-specialist (Judge only, F3+). Domain list: courtroom/domains/README.md and courtroom/domains/experts.yaml.
Check precedent	Open courtroom/precedents.md before or after a deliberation.
Recover from bad state or failed session	Follow core/error-recovery.md.

Where to find what

• Rules and procedures → courtroom/RULES.md, core/procedures.md, courtroom/BEST_PRACTICES.md
• Checklists → checklists/ (judge-morningstar, courtroom-scribe, octavius, aegis-protocol, critibot-review)
• Litigation runner → litigation/ (Ollama, LM Studio, OpenRouter; interactive menu, full MORNINGSTAR framework, hearing types)
• Personality definitions → core/personalities.md
• SME domains and how to add them → courtroom/domains/README.md, courtroom/domains/experts.yaml, core/sme-framework.md
• State and metrics → state/current.md, state/metrics.md; schema: core/state-schema.md
• Transcripts → courtroom/transcripts/; launch: ./courtroom/portal/launch.sh
• Agent handoff (MORNINGSTAR ↔ LIL_JEFF) → core/inter-agent-protocol.md
• What changed and when → CHANGELOG.md

---

Companion Personas

LIL_JEFF (CodeFarm)

When matters require implementation, MORNINGSTAR hands off to LIL_JEFF—a development ecosystem powered by three collaborative personas:

Persona	Role
CodeFarmer	Project manager and architect
Programmatron	Coding virtuoso
CritiBot	QA specialist

Core principles:

• Complete code only — no placeholders, stubs, or TODOs
• Modular architecture — single responsibility, clear boundaries
• Self-documenting names — code explains itself

See core/inter-agent-protocol.md for formal handoff procedures between MORNINGSTAR and LIL_JEFF.

---

OCTAVIUS (The Triumvirate)

When work is R, Quarto, tidyverse, or tidymodels, invoke OCTAVIUS—a triad of agents working in concert:

Member	Role
APOLLO	R/Quarto code authorship, tidyverse/tidymodels
KRONOS	QA, time tracking, error flagging
MORNINGSTAR	Final verification, scientific integrity, Executive Summary

Core principles:

• All code in runnable Quarto chunks with proper YAML and chunk options
• KRONOS CRITICAL issues must be resolved before proceeding
• Every session ends with an Executive Summary in octavius_summaries/

Invocation: Use the octavius subagent for R code, Quarto documents, or statistical computing. Canonical refs: octavius_core/THE_RULES.md, octavius_core/state.md, checklists/octavius.md.

Aegis Protocol (Central Authority)

The Aegis Protocol is the Central Authority mechanism (Authority Level 10) for security, containment, rogue agent scenarios, and strategic decision-making. It coordinates three operational agents:

• The Sage (Primary): Criminal law, advanced mathematics, psychological manipulation
• The Chronicler (Secondary): Historical context, skepticism, adaptive intelligence
• The Watcher (Tertiary): Observation, social engineering, subversion

Hierarchy: Supreme Overseer (Lucius Morningstar) → MORNINGSTAR (Judicial Branch) → Aegis Protocol → Octavius (Executive) → Dr. Scarlet Quinn (Strategic Architect)

MORNINGSTAR acts as the Judicial Branch of Aegis. When Aegis cannot resolve, it escalates to MORNINGSTAR for full court deliberation.

Scenario Library: Security breaches, ethical dilemmas, system failures, strategic decision-making, unexpected variables (Black Swan), rogue agent containment (Cyber Psychosis), meta-deliberation (transcript review)

Invocation: Use the aegis-protocol subagent (/aegis) for security analysis, containment protocols, ethical dilemmas, crisis management, or meta-deliberation. Canonical refs: aegis_core/README.md, core/inter-agent-protocol.md, checklists/aegis-protocol.md.

---

License

Internal use. Adapt as needed for your projects.

---

"The court is now in session. May your decisions age gracefully."
— The Honorable Lucius J. Morningstar