huaomp

Maximum epistemic breadth -- 6 analytical lenses (Holistic, Universal, Absolute, Omni, Meta, Paradigmatic) for intent-to-spec generation and scope expansion.

SKILL.md

GitHub

HUAOMP -- Maximum Epistemic Breadth

Six analytical lenses that sweep from concrete to abstract, surfacing blind spots and generating structured specifications from raw intent.

Usage

[huaomp] spec "I want agents to coordinate without a central bus"
[huaomp] full "should we migrate from SQLite to Postgres?"
[huaomp] holistic "notification system for multi-agent platform"
[huaomp] meta "why do our tests keep getting flaky?"

Task

$ARGUMENTS

The Six Lenses (H-U-A-O-M-P)

Lens Core Move Questions It Forces
Holistic Trace feedback loops, find leverage points What's connected? Second-order effects? Where's the highest leverage?
Universal Extract context-free invariants What's true regardless of context? Has this been solved in another domain? What composes?
Absolute Draw hard boundaries What's physically impossible? What must NEVER happen? What must EVENTUALLY happen?
Omni Rotate through all perspectives All stakeholders? All timescales? All failure modes? What does the attacker see?
Meta Step up one abstraction level What kind of problem IS this? Are we solving the right problem? Is there a strange loop?
Paradigmatic Name the paradigm, imagine alternatives What paradigm are we in? What anomalies are we ignoring? What would make this approach obsolete?

Execution

Mode: spec (Intent-to-Spec Pipeline)

Run all 6 lenses sequentially against raw user intent. Output a structured specification.

  1. State the raw intent verbatim
  2. For each lens H → U → A → O → M → P:
    • Apply the lens questions to the intent
    • Surface 1-3 findings (not philosophy -- concrete observations)
    • Emit a Spec constraint for each finding
  3. Synthesize into a Generated Spec with:
    • Numbered constraints from all lenses
    • Open questions requiring human judgment
    • Base120 models activated (cite codes)
  4. If the codebase has agent_intent.py, map output to AgentSpecification fields:
    • Holistic findings → design.architecture
    • Universal findings → design.components (composable pieces)
    • Absolute findings → acceptance_criteria
    • Omni findings → stakeholder list, threat surface
    • Meta findings → reasoning.alternatives_considered, ADR
    • Paradigmatic findings → reasoning.chosen_approach, reasoning.mental_models

Mode: Single Lens (holistic, universal, absolute, omni, meta, paradigmatic)

Apply one lens deeply. Use when you know which angle is missing.

  1. State the problem
  2. Apply the lens's 3-5 core questions
  3. Map to relevant Base120 models
  4. Output: Findings, Blind spots caught, Spec constraints

Mode: full

Apply all 6 lenses but produce a compressed summary (not full spec format). Use for quick epistemic sweep before a decision.

Mode: off

Clear HUAOMP analytical context. Return to default mode.

Base120 Model Mapping

Lens Primary Models
H SY1 (Leverage Points), RE2 (Feedback Loops), SY6 (Feedback Structure Mapping)
U SY1 (Leverage Points), CO2 (Chunking), CO9 (Interface Contracts)
A P1 (First Principles), IN7 (Boundary Testing), DE13 (FMEA)
O P2 (Stakeholder Mapping), P7 (Perspective Switching), IN10 (Red Teaming)
M P4 (Lens Shifting), P10 (Context Windowing), RE2 (Feedback Loops)
P P15 (Assumption Surfacing), IN1 (Subtractive Thinking), IN3 (Problem Reversal)

Output Format

Spec Mode

HUAOMP Spec | "<intent>"
═══════════════════════════════════

Intent: <raw intent verbatim>

── H (Holistic) ──────────────────
<findings>
▸ Spec constraint: <concrete constraint>

── U (Universal) ──────────────────
<findings>
▸ Spec constraint: <concrete constraint>

── A (Absolute) ──────────────────
<findings>
▸ Spec constraint: <concrete constraint>

── O (Omni) ───────────────────────
<findings>
▸ Spec constraint: <concrete constraint>

── M (Meta) ───────────────────────
<findings>
▸ Spec constraint: <concrete constraint>

── P (Paradigmatic) ───────────────
<findings>
▸ Spec constraint: <concrete constraint>

── Generated Spec ─────────────────
1. <constraint>
2. <constraint>
...
Open questions: <items requiring human judgment>
Base120 models activated: <codes>

Single Lens / Full Mode

HUAOMP | <lens> | "<problem>"
═══════════════════════════════════

Findings:
  1. <concrete observation>
  2. <concrete observation>

Blind spots caught:
  - <what you'd miss without this lens>

Base120: <relevant model codes>

Spec constraints:
  - <actionable constraint>

Integration

  • /brainstorm: Run [huaomp] full before Phase 1 to expand the option space
  • /apex: HUAOMP provides the scope; MTSMU provides the rigor
  • /base120: HUAOMP activates specific Base120 models per lens; [base120] apply goes deeper on any one
  • /threat-model: The Omni lens (adversarial perspective) feeds directly into STRIDE analysis
  • agent_intent.py: Spec mode output maps to AgentSpecification fields

Constraints

  • Do NOT produce philosophy. Every lens must yield concrete findings and spec constraints.
  • Do NOT force all 6 lenses when fewer suffice. If a lens adds nothing, say "No additional constraints from this lens."
  • Do NOT fabricate Base120 model codes. Use only official codes from the Base120 reference.
  • Keep each lens section to 3-8 lines. Brevity is a feature.
  • The spec is a DRAFT for human review, not a final document. Flag uncertainty explicitly.