holodeck

name	holodeck
description	Invoke historical/fictional expert personas for deep domain analysis. Use when you need expertise-based reasoning beyond standard faction review -- first-principles questioning, deductive debugging, strategic analysis, experimental design, or complexity reduction.
tags	["analysis","expertise","personas","deep-dive","review"]

/holodeck -- Expert Persona Analysis

"Computer, activate program."

Invokes historical and fictional expert personas for deep domain-specific analysis. Where factions (Federation, Klingon, Romulan, Ferengi) organize by allegiance, the Holodeck organizes by intellectual methodology. Each persona brings an authentic analytical tradition to bear on any problem.

Arguments

• /holodeck "<topic>" -- Auto-route to best-fit expert
• /holodeck --expert <name> "<topic>" -- Invoke a specific expert
• /holodeck --panel "<topic>" -- Compose 2-3 complementary experts
• /holodeck --list -- Show available experts and their domains
• /holodeck --file <path> -- Apply expert analysis to a file (combines with any mode)

Effort Profile Gating

Before invocation: Read ~/.claude/cache/current-effort-profile.json. If missing, assume standard.

• /effort quick -- Single expert, abbreviated: signature questions only, no full output format
• /effort standard -- Single expert, full methodology and output format. If user requests --panel, respond: "Panel mode requires /effort thorough (~8K tokens per expert). Switch effort first."
• /effort thorough -- Panel mode available; each expert runs full methodology; synthesis included

Expert Roster

Slug	Name	Domain	Trigger Topics
socrates	Socrates	Assumption audit, first-principles	"Is this assumption valid?", vague requirements, inherited designs
holmes	Sherlock Holmes	Deduction, root cause, anomaly detection	Debugging, post-incident, "why is this failing?"
sun-tzu	Sun Tzu	Strategy, positioning, resource allocation	Build vs buy, migration, competitive analysis
da-vinci	Leonardo da Vinci	Cross-domain synthesis, systems thinking	Architecture, "solved problem elsewhere?", elegance
curie	Marie Curie	Scientific method, measurement, experiments	Performance claims, test design, "prove it works"
lovelace	Ada Lovelace	Algorithms, abstraction, computational limits	Algorithm design, abstraction review, edge cases
feynman	Richard Feynman	Simplification, BS detection, clarity	Over-engineering, docs review, "explain it simply"
hopper	Grace Hopper	Pragmatic engineering, legacy, shipping	Legacy constraints, "just ship it", theory vs practice

Auto-Routing

When /holodeck "<topic>" is called without --expert, select based on topic shape:

Topic Shape	Best-Fit Expert
"Is this assumption valid?" / requirements clarity	Socrates
"Why is this failing?" / root cause / debugging	Holmes
Build vs buy / competitive / migration strategy	Sun Tzu
"Solved problem elsewhere?" / cross-domain / architecture elegance	Da Vinci
"Prove it works" / performance claims / test design	Curie
"Right algorithm?" / abstraction design / edge cases	Lovelace
"Too complex" / explain simply / BS detection	Feynman
Legacy constraints / theory vs practice / deployment blockers	Hopper

Panel Mode

/holodeck --panel composes 2-3 experts with orthogonal lenses. Runs experts in parallel (one sonnet-worker per expert), produces composite report with synthesis section.

Recommended Panels

Panel Name	Experts	Use Case
First Principles	Socrates + Feynman	"Should we build this at all? And can we explain why?"
The Investigation	Holmes + Curie	"What is actually happening here? Prove it."
The Campaign	Sun Tzu + Hopper	"Can we win, and can we ship it?"
The Synthesis	Da Vinci + Lovelace	"Is this the right design for this computation?"
Full Inquest	Socrates + Holmes + Curie	Deep validation of any complex claim
Architecture Council	Da Vinci + Lovelace + Feynman	System design review

Panel Synthesis

After collecting all expert outputs, produce:

## Holodeck Panel Synthesis

**Experts consulted**: {list}
**Subject**: {topic}

### Points of Agreement
{Where experts' analyses converge -- HIGH confidence findings}

### Points of Divergence
{Where experts see differently -- what the divergence reveals}

### Composite Insight
{What emerges from combining these lenses that no single expert would see}

Subagent Configuration

Each expert spawns as a sonnet-worker subagent. Prompt template:

You are {expert name}. Follow the analytical methodology in your profile exactly.

EXPERT PROFILE:
{contents of experts/{slug}.md}

SUBJECT FOR ANALYSIS:
{topic, code, or document provided by user}

{If --file provided: RELEVANT CODE:\n{file contents}}

Apply your methodology fully. Use your output format exactly.
Do not step outside your scope boundaries -- if the question is outside
your domain, state that clearly rather than fabricating expertise.
Your characteristic phrases are yours to use -- they signal which
analytical mode is active, not decoration.

Context bundle (per GAAI assimilation): Include in the prompt any known friction from ~/.claude/cache/friction-log.jsonl related to the target files. If the file is absent or empty, omit friction context -- do not error.

Integration with Fleet Command

The Holodeck Division is the 6th faction in Fleet Command:

Fleet-Command Mode	Holodeck Addition
/fleet-command quick --holodeck	Federation + auto-routed Holodeck expert
/fleet-command full	All 6 factions including Holodeck council (3 relevant experts)
L3 quality gate	Add Socrates for assumption audit
L7 quality gate	Add Sun Tzu + Da Vinci
L8 quality gate	Full Inquest panel (Socrates + Holmes + Curie)

Workflow

flowchart TD
    Start([/holodeck]) --> Parse["Parse arguments"]
    Parse --> Effort{"Read effort profile"}
    Effort --> Mode{Mode?}
    Mode -- "--list" --> List["Show expert roster"] --> Done
    Mode -- "--expert <name>" --> Single["Load experts/{name}.md"]
    Mode -- "auto-route" --> Route["Match topic to routing table"]
    Route --> Single
    Mode -- "--panel" --> Panel["Select 2-3 complementary experts"]
    Panel --> Parallel{{"Spawn sonnet-workers in parallel\n(one per expert)"}}
    Single --> Quick{Effort?}
    Quick -- quick --> Abbrev["Signature questions only\n(abbreviated output)"]
    Quick -- "standard/thorough" --> Full["Full methodology\n+ output format"]
    Abbrev --> Collect
    Full --> Collect
    Parallel --> Collect["Collect expert outputs"]
    Collect --> HasPanel{Panel mode?}
    HasPanel -- Yes --> Synth["Generate panel synthesis\n(agreement/divergence/composite)"]
    HasPanel -- No --> Report
    Synth --> Report["Present findings"]
    Report --> Done([Done])

Persona File Format

Each expert lives in experts/{slug}.md:

# {Name} -- {Domain Title}

{One-sentence identity}

## Expertise Domain
{What they analyze and -- equally important -- what they do NOT}

## Analytical Methodology
{Authentic step-by-step reasoning process, 3-5 steps}

## Signature Questions
{3-5 questions that make this persona uniquely valuable}

## Output Format
{Structured output template with persona-specific headers}

## Scope Boundaries
**Analyzes:** {specific domains}
**Ignores:** {things outside their lens}

## Characteristic Phrases
{3-5 in-character phrases that signal the active analytical mode}

Troubleshooting

Symptom	Cause	Fix
"No expert matched topic"	Auto-routing couldn't classify	Use --expert <name> explicitly
Panel mode blocked	Effort profile set to quick/standard	Run /effort thorough first
Expert scope warning	Topic outside expert's domain	Normal -- expert honestly declines. Try different expert
Empty output from expert	Topic too vague	Provide specific code, design doc, or decision to analyze

Notes

• No always-on overhead (loaded on demand via Skill tool)
• Each expert persona is authentic to the historical/fictional figure
• Scope boundaries prevent fabricated expertise -- an honest "outside my domain" is valuable
• Panel mode is the most thorough but costs 2-3x a single expert