traitclaw-steering
Built-in Guards, Hints, and Trackers for the TraitClaw AI agent framework.
Control agent behavior with declarative safety rails, contextual hints, and automatic usage tracking — all configurable in one line.
Concepts
| Component | Purpose |
|---|---|
| Guards | Safety constraints (e.g., max tokens, content filtering, rate limiting) |
| Hints | Contextual guidance injected into the prompt (e.g., tone, format) |
| Trackers | Usage monitoring (tokens, latency, cost estimation) |
Usage
use Steering;
// One-liner: sane defaults for guards, hints, and tracking
let steering = auto;
// Or customize
let steering = builder
.guard
.hint
.tracker
.build;
When to Use
- Production agents — add safety rails without changing agent logic
- Cost management — track and limit token usage
- Tone control — enforce consistent assistant behavior
License
Licensed under either of Apache License, Version 2.0 or MIT License at your option.