tonyflo79

deep-research

Comprehensive deep research system producing 1,000+ verbatim quotes across 6 buckets (Pain, Hope, Root Cause, Solutions Tried, Competitor Mechanism, Villain). Use when conducting market research for a new campaign, gathering audience intelligence, or building the evidence base for downstream copywriting. This is the most complex skill in the engine — it orchestrates multiple subagents across 4 layers with 57 microskills. Produces the research handoff package with market snapshots, avatar profiles, competitive landscape analysis, and categorized verbatim quotes. Trigger when users mention research, market research, audience research, gathering quotes, competitive analysis, or building the evidence base for a promotion. Requires Soul.md from Skill 00.

tonyflo79 0 1 Updated 2mo ago

Resources

17
GitHub

Install

npx skillscat add tonyflo79/ai-crush-vault/deep-research

Install via the SkillsCat registry.

SKILL.md

01: DEEP RESEARCH

1,000+ Verbatim Quotes Across 6 Buckets


PURPOSE

Build a comprehensive evidence base of real market language. This is not summary
research — it is verbatim quote mining from real people expressing real pain, hope,
frustration, and desire. The research handoff feeds every downstream skill.

Output: Research handoff package (market snapshot, avatar, competitive landscape, 1,000+ quotes)
Unlocks: Skill 02 (Proof Inventory) and all downstream skills


ARCHITECTURE

This skill is decomposed into 4 companion files (load only what you need):

  1. research-orchestrator.md — workflow orchestration (always load)
  2. research-layer-specs.md — per-layer execution specs (load relevant layer only)
  3. research-subagent-templates.md — model selection, personas, context templates
  4. research-output-protocol.md — output structure, checkpoints, session recap
  5. RESEARCH-ANTI-DEGRADATION.md — structural enforcement rules

NON-NEGOTIABLE THRESHOLDS

Bucket Minimum Quotes
Pain 300
Hope 250
Root Cause 200
Solutions Tried 150
Competitor Mechanism 100
Villain 75
TOTAL 1,000

Failure to meet any threshold = HALT. No conditional passes.