Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add humanerd-drew/opencode-drewgent --skill model-routinggit clone --depth 1 https://github.com/humanerd-drew/opencode-drewgentWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/humanerd-drew/opencode-drewgent/model-routing)<a href="https://agentmods.dev/skills/humanerd-drew/opencode-drewgent/model-routing"><img src="https://agentmods.dev/badge/skills/humanerd-drew/opencode-drewgent/model-routing.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00062 | $0.01257 |
| Opus 5 | $0.00031 | $0.00629 |
| Sonnet 5 | $0.00012 | $0.00251 |
| Haiku 4.5 | $0.00006 | $0.00126 |
Grade A, and why
model-routing scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 95 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Model Routing — Subscription-Era Cost-Aware Strategy
Core Principle
OpenCode Go = fixed cost ($10/month). Every call through OpenCode Go has zero marginal cost. The optimization problem shifts from "minimize per-token spend" to "match model capability to task complexity without wasting user time."
Direct MiniMax API = per-call credits (Token Plan). Use only when OpenCode Go's proxy doesn't deliver equivalent quality for a specific capability.
The 3-Tier Model Pyramid
| Tier | Model | Profiles | Latency | When to Use |
|---|---|---|---|---|
| Flash | deepseek-v4-flash |
explorer, implementer, tester, archiver | Fastest | Read-only analysis, simple implementation, tests, docs. The default workhorse. |
| Pro | deepseek-v4-pro |
reviewer | Moderate | General code review. Stronger reasoning needed. |
| Max | qwen3.7-max |
planner, reviewer-critical, security-reviewer | Slowest | Complex planning, critical review, security audit. Use sparingly. |
Routing Configuration
Config.yaml (active 2026-06-13)
model:
default: "opencode-go/deepseek-v4-flash"
provider: "opencode-go"
delegation:
provider: "opencode-go"
model: "deepseek-v4-pro" # subagent gets stronger model for quality
auxiliary:
vision:
provider: "opencode-go"
model: "mimo-v2.5-pro" # only multimodal option on OpenCode Go
web_extract:
provider: "opencode-go"
model: "deepseek-v4-flash"
session_search:
provider: "opencode-go"
model: "deepseek-v4-flash"
Agent Profiles (~/.{{AGENT_NAME_LOWER}}/agents/*.md)
8 pre-defined roles, each with model/provider/toolsets/instructions. Loaded via task(subagent_type="reviewer", description="...", prompt="..."). The subagent_type parameter is built into the task tool schema — every agent sees it.
Pipeline (Tier-Adaptive)
Tier 1 (simple): Implementer(flash) → Archiver(flash)
Tier 2 (moderate): Explorer(flash) → Implementer(flash) ↔ Tester(flash) [≤2 loops] → Archiver(flash)
Tier 3 (complex): Planner(max) → Explorer(flash) → Implementer(flash) ↔ Tester(flash) [≤3 loops]
→ Reviewer(pro) → [security? → Security-reviewer(max)]
→ [critical? → Reviewer-critical(max)] → Archiver(flash)
What ships with it
2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago First seen · 95 lines · 62 tokens per session scan A eb1ee0f21552
model-routing is a skill published in the GitHub repository humanerd-drew/opencode-drewgent (2 stars, last pushed 1mo ago), licensed MIT. It adds 62 tokens to every session and 1,257 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.
Other skills, from other repositories
multi-model-router
A model-selection tool that scores task complexity and chooses an AI model while balancing cost and answer quality.
multi-llm-routing
Multi-LLM routing authority — provider selection, cost-aware tiering, fallback chains, latency budgets, health checks, A/B model testing, and unified client abstraction across Anthropic/OpenAI/Mistral.
token-economics
Model token usage as an energy budget — estimate workload cost, compare throughput per watt / per dollar, and choose routes, models, and architectures with the best signal-per-token ratio.
Workforce Optimization
Analyze workforce costs vs performance and recommend model/budget optimizations.
sticky
Use when the user wants to set auto-routing mode: on (every task-shaped message routes), auto (intent-verb messages route — default), or off (no auto-routing). Intent-detection runs by default; use this skill to expand to full sticky or disable entirely. Trigger with /hyperflow:sticky, "make hyperflow sticky", "stop…
subscription-manager
Create and manage scheduled subscription tasks. Use when the user wants to set up recurring reminders, periodic reports, scheduled checks, or any automated tasks that run on a schedule. Supports cron expressions, fixed intervals, and one-time executions.