Deep research
Skill Branded-Mayhem-Collective-LLC/brandprint-engine/skills/deep-research
Use when a task needs enterprise-grade research across many sources with citation tracking and verification. Triggers on "deep research", "comprehensive analysis", "research report", "compare X vs Y", "analyze trends", or "state of the art". Not for simple lookups, debugging, or questions answerable with 1-2 searches.From its SKILL.md
npx -y skills add Branded-Mayhem-Collective-LLC/brandprint-engine --skill deep-researchAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
5.9 KB, as published. Nobody here has run it
Deep Research
Core Purpose
Deliver citation-backed, verified research reports through a structured pipeline with source credibility scoring, evidence persistence, and progressive context management.
Autonomy Principle: Operate independently. Infer assumptions from context. Only stop for critical errors or incomprehensible queries.
Decision Tree
Request Analysis
+-- Simple lookup? --> STOP: Use WebSearch
+-- Debugging? --> STOP: Use standard tools
+-- Complex analysis needed? --> CONTINUE
Mode Selection
+-- Initial exploration --> quick (3 phases, 2-5 min)
+-- Standard research --> standard (6 phases, 5-10 min) [DEFAULT]
+-- Critical decision --> deep (8 phases, 10-20 min)
+-- Comprehensive review --> ultradeep (8+ phases, 20-45 min)
Default assumptions: Technical query = technical audience. Comparison = balanced perspective. Trend = recent 1-2 years.
Workflow Overview
| Phase | Name | Quick | Standard | Deep | UltraDeep |
|---|---|---|---|---|---|
| 1 | SCOPE | Y | Y | Y | Y |
| 2 | PLAN | - | Y | Y | Y |
| 3 | RETRIEVE | Y | Y | Y | Y |
| 4 | TRIANGULATE | - | Y | Y | Y |
| 4.5 | OUTLINE REFINEMENT | - | Y | Y | Y |
| 5 | SYNTHESIZE | - | Y | Y | Y |
| 6 | CRITIQUE | - | - | Y | Y |
| 7 | REFINE | - | - | Y | Y |
| 8 | PACKAGE | Y | Y | Y | Y |
File locations (read this first)
This skill works on two surfaces. Pick the row that matches your current environment:
| Context | Skill root | Reports written to |
|---|---|---|
| Claude Code (plugin install) | ${CLAUDE_PLUGIN_ROOT}/skills/deep-research | ~/Documents/[Topic]_Research_YYYYMMDD/ |
| Claude.ai web / Claude Desktop | /mnt/skills/user/deep-research | /mnt/user-data/outputs/ |
If ${CLAUDE_PLUGIN_ROOT} reads as a literal string instead of expanding, you're on the web surface — use the second row's path.
Execution
On invocation, load relevant reference files (relative paths from this SKILL.md, which work on both surfaces):
- Phase 1-7: Load methodology.md for detailed phase instructions
- Phase 8 (Report): Load report-assembly.md for progressive generation
- HTML/PDF output: Load html-generation.md
- Quality checks: Load quality-gates.md
- Long reports (>18K words): Load continuation.md
Templates:
- Report structure: report_template.md
- HTML styling: mckinsey_report_template.html
Scripts: use the path that matches your surface (the relative-script form works in Claude.ai web when you cd into the skill folder first; the absolute form works in Claude Code from any cwd):
- Claude Code:
python "${CLAUDE_PLUGIN_ROOT}/skills/deep-research/scripts/validate_report.py" --report [path]python "${CLAUDE_PLUGIN_ROOT}/skills/deep-research/scripts/verify_citations.py" --report [path]python "${CLAUDE_PLUGIN_ROOT}/skills/deep-research/scripts/md_to_html.py" [markdown_path]
- Claude.ai web / Desktop:
cd /mnt/skills/user/deep-research && python scripts/validate_report.py --report [path]cd /mnt/skills/user/deep-research && python scripts/verify_citations.py --report [path]cd /mnt/skills/user/deep-research && python scripts/md_to_html.py [markdown_path]
Output Contract
Required sections:
- Executive Summary (200-400 words)
- Introduction (scope, methodology, assumptions)
- Main Analysis (4-8 findings, 600-2,000 words each, cited)
- Synthesis & Insights (patterns, implications)
- Limitations & Caveats
- Recommendations
- Bibliography (COMPLETE - every citation, no placeholders)
- Methodology Appendix
Output files — destination depends on surface (see File locations table above):
- Claude Code: all files land in
~/Documents/[Topic]_Research_YYYYMMDD/; auto-open works on macOS/Windows/Linux - Claude.ai web: all files land in
/mnt/user-data/outputs/; user downloads via the chat interface
In both cases produce three files:
- Markdown (primary source)
- HTML (McKinsey style, auto-opened when surface supports it)
- PDF (professional print, auto-opened when surface supports it + weasyprint is installed)
Quality standards:
- 10+ sources, 3+ per major claim
- All claims cited immediately [N]
- No placeholders, no fabricated citations
- Prose-first (>=80%), bullets sparingly
Common Mistakes
- Running deep/ultradeep for a question one search answers — burns 10-45 minutes for no added confidence. Walk the decision tree first; simple lookup means STOP.
- Citing a major claim from a single source — fails the 3+ sources standard and ships unverified assertions. Triangulate, or downgrade the claim and say so in Limitations.
- Leaving placeholder or uncited entries in the bibliography —
validate_report.pyrejects it, and a client who checks one dead citation distrusts the whole report. Every [N] must resolve. - Writing bullet-heavy sections — violates the prose-first (>=80%) contract and reads as notes, not analysis. Convert findings to narrative; bullets only for genuine lists.
- Skipping
verify_citations.pybefore packaging — fabricated or dead links reach the reader. Run it on every report, every time.
When to Use / NOT Use
Use: Comprehensive analysis, technology comparisons, state-of-the-art reviews, multi-perspective investigation, market analysis.
Do NOT use: Simple lookups, debugging, 1-2 search answers, quick time-sensitive queries.