Progressive disclosure for what AI agents produce — Summary → Analysis Collection → Dossiers.

The Artifact Pyramid is a structured methodology for organizing AI agent research outputs across three layers of increasing depth. Just as progressive disclosure governs how we feed agents context, the Artifact Pyramid governs what they produce — enabling downstream agents and humans to consume at the depth they need.

flowchart TD
    subgraph Layer1["Layer 1 — Summary 🎯"]
        direction LR
        S1[Key findings] --- S2[Implications] --- S3[Links to analysis]
    end

    subgraph Layer2["Layer 2 — Analysis Collection 🧩"]
        direction LR
        A1[Market] --- A2[Competitive] --- A3[Technical] --- A4[Risk]
    end

    subgraph Layer3["Layer 3 — Detailed Dossiers 📦"]
        direction LR
        D1[Source excerpts] --- D2[Transcripts] --- D3[Raw data] --- D4[Methodology]
    end

    Layer1 --> Layer2
    Layer2 --> Layer3

    style Layer1 fill:#1a1a2e,stroke:#e94560,color:#fff
    style Layer2 fill:#16213e,stroke:#0f3460,color:#fff
    style Layer3 fill:#0f3460,stroke:#533483,color:#fff

Current AI agent workflows use progressive disclosure on the input side (metadata → instructions → resources) but produce flat, monolithic outputs on the output side. The Artifact Pyramid fixes this asymmetry, making agent outputs:

• Independently consumable at every layer of fidelity
• Bidirectionally traceable from published claim back to source
• Pipeline-friendly for multi-agent research workflows
• Quality-gated at each transformation step

# Check pyramid health of a research project
scripts/pyramid-status.sh ./my-project

# Extract candidate atoms from a source file
scripts/extract-atoms.py source.txt --source-id paper-001 --domain scaling-laws

# Scaffold a new research project
cp assets/pyramid-template.md ./my-project/00-index.md

artifact-pyramids/
├── SKILL.md                    # Agent Skills-compliant skill (loadable by AI agents)
├── README.md                   # This file
├── LICENSE                     # MIT
├── scripts/
│   ├── pyramid-status.sh       # Audit a project directory for structural coverage
│   └── extract-atoms.py        # Extract atomic claims from source text
├── references/
│   ├── artifact-pyramid-framework.md   # Full conceptual foundation
│   ├── pipeline-stages.md              # Detailed layer definitions and navigation format
│   ├── quality-gates.md                # Verification criteria at each layer
│   └── synthetic-example.md            # Complete walked-through example (synthetic data)
└── assets/
    ├── pyramid-template.md             # Project scaffold template
    └── artifact-inventory.md           # Cross-layer tracking template

This repo ships as an Agent Skills-compliant skill. To load it in Hermes Agent:

git clone https://github.com/groktopus/artifact-pyramids ~/.hermes/skills/artifact-pyramids

Then any session with the skill loaded can call skill_view(name='artifact-pyramids') to activate it.

MIT