Babysitter agent-booster

WASM-based instant code transforms for simple tasks, achieving 352x speedup over LLM inference with zero cost.

install

source · Clone the upstream repo

git clone https://github.com/a5c-ai/babysitter

Claude Code · Install into ~/.claude/skills/

T=$(mktemp -d) && git clone --depth=1 https://github.com/a5c-ai/babysitter "$T" && mkdir -p ~/.claude/skills && cp -r "$T/library/methodologies/ruflo/skills/agent-booster" ~/.claude/skills/a5c-ai-babysitter-agent-booster && rm -rf "$T"

manifest: library/methodologies/ruflo/skills/agent-booster/SKILL.md

Agent Booster

Overview

WASM-compiled code transformation engine for simple, well-defined tasks. Bypasses LLM inference entirely for pattern-matched transforms, achieving sub-millisecond execution at zero cost.

When to Use

Simple, deterministic code transforms
High-frequency repetitive modifications
When latency is critical (<1ms requirement)
Cost-sensitive batch operations

Supported Transforms

Transform	Description	Example
`var-to-const`	Modernize variable declarations	`var x = 1` -> `const x = 1`
`add-types`	Insert TypeScript annotations	`function f(x)` -> `function f(x: string)`
`add-error-handling`	Wrap in try/catch	Bare calls -> try/catch blocks
`async-await`	Convert Promise chains	`.then().catch()` -> `async/await`
`extract-function`	Extract code blocks	Inline code -> named function
`inline-variable`	Inline single-use variables	Remove intermediate vars
`add-jsdoc`	Generate documentation	Bare functions -> JSDoc comments

Performance

Execution: <1ms per transform
Cost: $0 (no LLM invocation)
Speedup: 352x compared to LLM inference
Confidence threshold: >90% pattern match required

Agents Used

```
agents/coder/
```
- Fallback for unmatched patterns

Tool Use

Invoke via babysitter process:

methodologies/ruflo/ruflo-task-routing