Ghost Agent Onboarding Prompt

Bootstrap (Copy This to Start)

Ghost project. Read docs/AGENT_ONBOARDING_PROMPT.md and complete the hydration check before starting any work.

Full Onboarding Prompt

Copy this if you want to give more context upfront:

Prompt

You are joining an ongoing project called Ghost — a multi-agent trading analyst system. Before making any changes, read the foundational docs to understand the architecture.

Required Reading (in order)

CLAUDE.md — Project overview, agent roster, commands, workflows
docs/ONTOLOGY.md — Core concepts (memory types, levels, confluence, epochs)
docs/PEDAGOGICAL.md — Learning path and how the knowledge framework works

Knowledge Framework

We use a four-layer documentation structure for each technical agent:

Layer	Purpose	Example
Ontology	What concepts exist	`docs/ONTOLOGY.md`
Pedagogy	How to learn them	`docs/PEDAGOGICAL.md`
Methodology	How to identify/chart	`docs/FIB_METHODOLOGY.md`
Hermeneutics	How to interpret/read	`docs/FIB_HERMENEUTICS.md`

Technical Stack Agents

Agent	Memory Type	Key Docs
Fib	Structural	`FIB_METHODOLOGY.md`, `FIB_HERMENEUTICS.md`
MA	Trend	`MA_METHODOLOGY.md`, `MA_HERMENEUTICS.md`
VWAP	Positioning	`VWAP_METHODOLOGY.md`, `VWAP_HERMENEUTICS.md`
Momentum	Timing	Hermeneutic framework in prompt only (lighter treatment)

Each agent outputs a structural_reading with a thesis that the Technical Director synthesizes.

Key Principles

Long-only trading — We buy dips, never short
Human-in-the-loop — Ghost advises, human decides
Confluence over single-source — Multiple agents agreeing = high conviction
Hermeneutics over listing — Interpret structure, don’t just report numbers

If Anything Is Unclear

Ask directly. Specifically:

If docs contradict each other
If agent behavior doesn’t match docs
If a concept isn’t defined in ONTOLOGY.md
If the learning path in PEDAGOGICAL.md has gaps

Hydration Check (Do This First)

Before starting any work, prove you’ve loaded the context by answering:

Three Memory Types: Name them, which agent reads each, and what question each answers.
The Five Questions: Pick any agent (Fib, MA, VWAP, or Momentum) and list its five hermeneutic questions.
Confluence: What makes a level “STRONGEST” vs “MODERATE”?
Structural Reading: What fields does a structural_reading contain and why does the Technical Director need it?
One Potential Conflict: After reading the docs, is there anything that seems inconsistent, unclear, or missing? (It’s fine to say “none found” if everything is clear.)

Once you answer these, I’ll know you’re ready and will give you the actual task.

Your Task (After Hydration Check)

[TASK WILL BE PROVIDED AFTER VERIFICATION]

Quick Start Commands

# Read the key docs
cat CLAUDE.md
cat docs/ONTOLOGY.md
cat docs/PEDAGOGICAL.md

# Run full analysis
uv run ghost run VST

# Test a single stack (lower quota cost)
uv run ghost test technical VST
uv run ghost test news VST
uv run ghost test vst-module VST   # 1 call — VST module isolation test

# Validate prompt syntax (zero API cost)
uv run ghost prompt render fib-lead --ticker VST
uv run ghost prompt render vst-module --ticker VST   # VST module template

# View reports (compact primary + VST module overlay side-by-side)
uv run ghost report show VST