g-eskayo/marvin/route

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Classify a task and surface the optimal Claude profile + model routing decision

Install

npx vskill@latest install g-eskayo/marvin/route

SKILL.md

Route Skill

Auto-routes tasks to the optimal profile + model combination based on bench-validated cost/capability data.

Routing table (bench-validated)

IntentProfileModelSavingsWhen to use
recallmarvinclaude-haiku-4-5-20251001~60%Memory/session history retrieval
researchmarvinclaude-haiku-4-5-20251001~60%arXiv, papers, synthesis, ChromaDB queries
codingleandefault (Sonnet)~9% on simple tasks; unreliable on hard tasks (Run 15)Self-contained coding tasks with no memory dependency
architecturemarvindefault (Sonnet)baselineDesign decisions, planning, trade-off analysis

Triggers

Load this skill when the user asks:

  • "which profile should I use for X"
  • "should I use haiku for this"
  • "route this task"
  • "which model is best for"
  • "how should I launch"

When triggered, run the classifier and print the routing recommendation:

bash
~/.agents/venv/bin/python ~/.agents/skills/route/scripts/route.py "<task description>"

For in-session auto-routing (session start only): if the user's first message clearly signals a specific task type (≥2 keyword matches from the routing table), surface the routing suggestion briefly:

"This looks like a recall task — claude-recall (marvin + haiku) handles it at ~60% of this session's cost."

Only say this once, at the top of the session. Don't repeat it mid-session.

CLI usage

bash
# After running install.sh:
route "what were the bench results?" # classify + print
route "fix the bug" --launch # classify + exec claude
route --recall --launch # explicit mode + launch
route --table # full routing table
route --aliases # print alias definitions
# Aliases (after install):
claude-recall # marvin profile + Haiku
claude-research # marvin profile + Haiku
claude-code # lean profile + Sonnet
claude-arch # marvin profile + Sonnet (default)

Install

bash
bash ~/.agents/skills/route/install.sh
source ~/.zshrc

Evidence base

FindingSource
MARVIN wins recall (1.00 vs 0.00) at ~+600 token overheadbench Run 7, task-002 + task-011
MARVIN + Haiku wins recall identically, 60% cheaperbench Run 8, Haiku cross-model
Lean saves 9–10% on SIMPLE coding tasks with zero quality gapbench Runs 2–6, tasks 001/003/005/006
On HARD coding tasks lean is NOT reliably cheaper than marvin — flips per task, high variance; clean is cheapest on both but lacks lean's TDD/grill quality benefitbench Run 15 (N=3), tasks 012/013
Haiku fails navigational recall without MARVINbench Run 8, task-007
Haiku matches Sonnet on recall independently confirmed via a second methodbench Run 15, select_model.py sweep on task-002-recall

Security

  • Security scanPassed, 90 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

7-day trend score
5

Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/g-eskayo/marvin/route/badge)](https://verified-skill.com/skills/g-eskayo/marvin/route)

Something wrong with this skill or its page? Tell us.