← Back to Skills Marketplace
SwitchBoard
by
gigabit-eth
· GitHub ↗
· v1.0.2
714
Downloads
2
Stars
5
Active Installs
3
Versions
Install in OpenClaw
/install router
Description
Cost-optimize AI agent operations by routing tasks to appropriate models based on complexity. Use this skill when: (1) deciding which model to use for a task...
Usage Guidance
This skill appears coherent and does what it says: pick cost-appropriate OpenRouter models. Before installing: (1) confirm you trust the skill source (homepage/repo in metadata point to a third-party project — verify ownership if you need provenance), (2) only store your OPENROUTER_API_KEY in ~/.openclaw/openclaw.json with strict permissions (600) as suggested, (3) avoid routing sensitive secrets/PII to free or unmoderated models (the skill already warns about this), (4) remember this skill will cause the agent to call openrouter.ai endpoints at runtime, so review OpenRouter’s privacy/policy for any chosen model, and (5) because the package is instruction-only there is no bundled executable, but the agent will act on its routing rules — if you want stricter control, limit the skill to manual invocation rather than autonomous agent flows.
Capability Analysis
Type: OpenClaw Skill
Name: router
Version: 1.0.2
The OpenClaw AgentSkills bundle is benign. It transparently declares its purpose of cost-optimizing AI agent operations by routing tasks to different models via the OpenRouter API. The `SKILL.md` provides clear instructions to the agent for model selection without any malicious prompt injection attempts. All required API keys (`OPENROUTER_API_KEY`) and network endpoints (`https://openrouter.ai/api/*`) are explicitly declared in `manifest.json` and `skill.json`. The skill contains no custom executable code, binaries, or external dependencies, and responsibly warns users about potential privacy implications of third-party models it might route to.
Capability Assessment
Purpose & Capability
Name/description (model routing, cost optimization) align with requested artifacts: an OpenRouter API key, model catalog files, and routing logic in SKILL.md. The large references/openrouter-models.* files are consistent with a routing/reference skill.
Instruction Scope
SKILL.md is instruction-only and tells the agent to read/store the OpenRouter key in ~/.openclaw/openclaw.json and to route tasks based on complexity. Instructions remain within routing scope and explicitly warn not to send secrets to unmoderated/free models. Note: it will direct runtime calls to openrouter.ai if used, and the skill relies on model metadata in the included reference files.
Install Mechanism
No install spec or code is provided (instruction-only), so nothing is downloaded or installed by the skill itself. This minimizes disk-write/exec risk.
Credentials
Only one credential is declared (OPENROUTER_API_KEY) and its use is justified by the skill's purpose. The manifest documents the config path (~/.openclaw/openclaw.json) and recommended file perms (600). No unrelated secrets or multiple credentials are requested.
Persistence & Privilege
The skill is not always-enabled and is user-invocable. It does not request elevated persistence or modify other skills; declared persistence is limited to the user's OpenClaw config file for the API key.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install router - After installation, invoke the skill by name or use
/router - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.2
- Added manifest.json and skill.json files for improved metadata and compatibility.
- Updated documentation to include a privacy note about model logging and a warning not to route sensitive data through free/unmoderated models.
- No changes to routing logic or classification guidance.
v1.0.1
- Added: OpenRouter API KEY Prequisite
- Tighter documentation integration
- Optimized control flow logic
v1.0.0
Initial release: introduces a structured approach for cost-efficient AI model selection by routing tasks based on complexity.
- Defines a three-tier model hierarchy (Routine, Moderate, Complex) linked to cost ranges and suitable tasks.
- Provides classification rules, example model assignments, and decision algorithms for task-to-model routing.
- Outlines best practices and anti-patterns for main sessions, sub-agents, and automated tasks.
- Includes integration examples for common agent frameworks.
- Details the cost impact and recommended usage splits for optimal budget management.
Metadata
Frequently Asked Questions
What is SwitchBoard?
Cost-optimize AI agent operations by routing tasks to appropriate models based on complexity. Use this skill when: (1) deciding which model to use for a task... It is an AI Agent Skill for Claude Code / OpenClaw, with 714 downloads so far.
How do I install SwitchBoard?
Run "/install router" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is SwitchBoard free?
Yes, SwitchBoard is completely free (open-source). You can download, install and use it at no cost.
Which platforms does SwitchBoard support?
SwitchBoard is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created SwitchBoard?
It is built and maintained by gigabit-eth (@gigabit-eth); the current version is v1.0.2.
More Skills