@dzhechkov/scout

Multi-source ecosystem intelligence — scans 5 sources for new agent-skill projects, detects formats, scores relevance, and generates actionable intelligence reports.

Sources (5)

Usage

npm i -g @dzhechkov/harness-cli

# Quick scan — radar mode (all 5 sources)
dz scout

# Deep analysis — downloads SKILL.md, finds gaps, recommends integration
dz scout --deep

# Custom topics (GitHub-only)
dz scout --topics mcp-server,ai-agent

# Only recent
dz scout --since 2026-05-01

What it does

Radar mode (dz scout)

1. Scans 5 sources in parallel with deduplication
2. Detects skill formats: SKILL.md, plugin.json, .claude/skills/, MCP manifests
3. Scores relevance: format (40%) + stars (30%) + recency (20%) + novelty (10%)
4. Compares against our 24 packages — finds skills we don't have
5. Reports with recommendations: integrate / monitor / skip
6. Shows source provenance — which source found each result

Deep analyst mode (dz scout --deep)

For top-scored repos (score ≥50), goes further:

1. Downloads SKILL.md from each repo, parses frontmatter
2. Finds closest match in our inventory by keyword overlap
3. Explains the delta — what the found skill adds that ours doesn't
4. Recommends integration path:
   • canonicalize — novel skill → new @dzhechkov/skills-* pack
   • merge — similar to existing → add unique features to ours
   • new-preset — novel → add to preset or new pack
   • skip — already in inventory
5. Gap analysis — trending categories we lack

Example output

Scanning 5 sources (GitHub + npm + HN + MCP Registry + Glama)...

Sources: github: 50, npm: 42, hackernews: 30, mcp-registry: 20

# Scout Intelligence Report
Total found: 123 | Scanned: 123

## 🟢 Integrate (high relevance)
| Repo | Stars | Formats | Score | Novel skills |
|------|-------|---------|-------|-------------|
| fallow-rs/fallow-skills | 58 | agentskills-io | 88 | dead-code, codebase-hygiene |

## 📊 Harness Gap Analysis
| Category | Frequency | Recommendation |
|----------|-----------|---------------|
| mcp | 10 repos | Create @dzhechkov/skills-mcp |

Programmatic API

import { scanAllSources, deepAnalyze, generateReport } from '@dzhechkov/scout';

// Multi-source scan
const { results, totalBySource } = await scanAllSources({ token: process.env.GITHUB_TOKEN });

// Generate report
const report = generateReport({ repos: results, ... });

// Deep analysis on top results
const deep = await deepAnalyze(results, { token });

// Individual sources
import { scanNpm, scanHN, scanMcpRegistry, scanGlama } from '@dzhechkov/scout';

Status

v0.4.0 — published on npm. Part of DZ Harness Hub.