dz scout

dz scout scans 9 sources in parallel for new AI agent skills, MCP servers, and community skill packages. It generates an intelligence report with gap analysis and integration recommendations. Use --deep for SKILL.md analysis, --report to print the report, --diff to compare against the last scan, and --output to save the report.

Basic Usage

# Quick scan (all 9 sources)
dz scout

# Scan repos updated since a date
dz scout --since 2026-05-01

# Restrict to specific topics
dz scout --topics mcp,web3,testing

# Deep mode: SKILL.md parsing + gap analysis
dz scout --deep

# Save report to a markdown file
dz scout --output ./reports/scout-report.md

# Print the report / diff against last scan
dz scout --report
dz scout --diff

The 9 Sources

dz scout queries these sources in parallel:

#	Source	What It Scans
1	GitHub	agent-skills, claude-code, mcp-server topics
2	npm Registry	Packages with agent-skills, claude-skill, mcp-server keywords
3	Hacker News	HN Algolia posts mentioning AI skills, MCP, agent workflows
4	MCP Registry	Official MCP server registry
5	Glama	Glama MCP server directory
6	OSSInsight	Trending repositories by star velocity
7	Smithery	Smithery MCP marketplace
8	Semantic Scholar	Academic papers with citation counts and impact metrics
9	arXiv	Research papers: multi-agent systems, tool use, skill learning

Scout compares discovered packages against the 31 DZ Harness Hub packages to find skills we don’t yet have.

Example Output

DZ HARNESS HUB — Scout Report
──────────────────────────────
Sources scanned: 9
Results found:   231
New since last:  34

TOP RESULTS BY SOURCE
─────────────────────
GitHub          87 repos   top: nexu-io/open-design (57K★)
npm             42 pkgs    top: @lythos/agent-skills-core
MCP Registry    31 tools   top: brave-search, filesystem, git
Smithery        18 tools   top: context7, exa-search
Glama           14 tools   top: notion, github
OSSInsight       9 repos   top: star-velocity > 500/week
Semantic Scholar 8 papers  top: "Multi-Agent Skill Composition"
arXiv            8 papers  top: "Tool-Augmented LLMs" (2026-06)
HN Algolia       7 posts   top: "Show HN: Open MCP Skills Hub"

GAP ANALYSIS
─────────────────────────────
ai-agents        15 repos  integrate → suggest dz import-ecc
mcp              10 repos  monitor → extend skills-mcp preset
observability     8 repos  monitor → add to roadmap

Deep Mode (--deep)

In deep mode, dz scout fetches the SKILL.md (or equivalent) from each GitHub result and runs additional analysis:

• SKILL.md parsing — extracts title, description, version, tags, trust_tier
• Gap scoring — compares against existing skills to measure novelty
• Integration recommendation — one of: integrate, monitor, skip

Relevance is scored by format (40%) + stars (30%) + recency (20%) + novelty (10%). Recommendations: integrate (score ≥ 70), monitor (40–69 with ≥ 50 stars), or skip.

# Deep scan example
dz scout --deep

# Sample deep analysis output:
#  nexu-io/open-design  57K★  → integrate (UI/UX design workflows)
#  dotnet/skills        3K★   → monitor (5 skills, partial overlap)
#  ms/power-platform    330★  → skip (enterprise-only, not portable)

Memory Integration

dz scout stores its results so future runs can diff against past reports:

# First run stores the result set
dz scout

# Later, compare against the last scan
dz scout --diff
# → "New since last: 34"

Options

Flag	Description
--topics <list>	Comma-separated topics to focus the scan
--since <date>	Only include repos/packages updated since this date (e.g. 2026-05-01)
--deep	Enable SKILL.md parsing and gap analysis
--output <file>	Save report to a markdown file
--report	Print the intelligence report
--diff	Compare against the last saved scan

Examples

# Scout recent repos and save the report
dz scout --since 2026-05-01 --output ./scout-reports/report.md

# Deep analysis focused on MCP and web3
dz scout --deep --topics mcp,web3

# Diff against the previous scan
dz scout --diff

Related Commands

• dz import-ecc — import skills discovered by scout
• dz auto-canonicalize — generate create-skill commands from GitHub repos
• dz create-skill — scaffold a new skill from a scouted source
• dz recommend — task-based skill recommendations