Smart Skill Router

Smart Skill Router

Architecture

User Request
      |
      v
+------------------+
| Query Analyzer   | <-- Extract intent, domain, keywords
+------------------+
      |
      v
+------------------+
| Skill Matcher    | <-- skills_index.json (instant lookup)
+------------------+
      |
      v
+------------------+
| Ranking Engine   | <-- Score: relevance + tags + priority + deps
+------------------+
      |
      v
+------------------+
| Dependency       | <-- skills_graph.json (load prerequisites)
| Resolver         |
+------------------+
      |
      v
+------------------+
| Bundle Detector  | <-- If query spans a bundle, activate it
+------------------+
      |
      v
Execute Top Skills (max 3-5 concurrently)

Quick Usage

From Python

import sys
sys.path.insert(0, r'C:\Users\Mamoud\.claude\skills')
from skill_router import find_skill, route_query, resolve_deps, get_bundle

# Search by keyword
results = find_skill("liquidity")

# Search by tags
results = find_skill(tags=["orderflow", "liquidity"])

# Full routing with scoring
top_skills = route_query("analyze liquidity trap on EURUSD H1")

# Dependency chain
deps = resolve_deps("ict-smart-money")

# Get a bundle
bundle = get_bundle("scalping")

From CLI

python C:\Users\Mamoud\.claude\skills\skill_router.py search "breakout strategy"
python C:\Users\Mamoud\.claude\skills\skill_router.py route "analyze smart money on gold"
python C:\Users\Mamoud\.claude\skills\skill_router.py deps ict-smart-money
python C:\Users\Mamoud\.claude\skills\skill_router.py bundle scalping
python C:\Users\Mamoud\.claude\skills\skill_router.py stats
python C:\Users\Mamoud\.claude\skills\skill_router.py info risk-and-portfolio

Scoring Formula

score = relevance + tag_match + dependency_bonus + priority - overlap_penalty

Where:
  relevance       = keyword match in name (+10) or description (+5)
  tag_match       = keyword match in tags (+8 per tag)
  dependency_bonus = +3 if a matched skill depends on this one
  priority        = skill priority value (1-10)
  overlap_penalty = -2 if another higher-scoring skill covers same topic

Predefined Bundles

Auto-Selection Protocol

When the user makes a request:

1. Extract keywords from the request
2. Run route_query() to get ranked skills
3. Check if a bundle applies — if 3+ skills from one bundle match, activate the full bundle
4. Resolve dependencies for the top skills
5. Execute in order: dependencies first, then primary skills
6. Record usage via skill_health.py

Files