JPeetz/agent-skills / skills/development/model-routing-cost-optimizer
model-routing-cost-optimizer
VerifiedUse when routing tasks to the cheapest fitting model tier.
| File | Kind | Size |
|---|---|---|
| CHANGELOG.md | markdown | 566 B |
| LICENSE | other | 1.1 KB |
| SKILL.md | skill md | 6.7 KB |
| evals/evals.json | config | 3.2 KB |
| references/model-pricing.md | markdown | 1.9 KB |
| references/sources.md | markdown | 914 B |
| scripts/classify_task.py | script | 2.3 KB |
scripts/classify_task.py
View on GitHub#!/usr/bin/env python3
"""Classify a task description into a cost tier: ROUTINE / MODERATE / COMPLEX.
A small, stdlib-only classifier used by the model-routing cost optimizer. It
reads a task description plus optional complexity keywords and returns the
recommended tier mapping (ROUTINE -> Tier 1, MODERATE -> Tier 2, COMPLEX ->
Tier 3). It is a heuristic aid, not a substitute for the vision/escalation
overrides in SKILL.md.
Usage:
python3 scripts/classify_task.py "<task description>" [keyword...]
Examples:
python3 scripts/classify_task.py "read the config file and log it"
python3 scripts/classify_task.py "debug this crash and fix the cause" debug
"""
from __future__ import annotations
import re
import sys
ROUTINE = "ROUTINE"
MODERATE = "MODERATE"
COMPLEX = "COMPLEX"
ROUTINE_SIGNALS = {
"read", "fetch", "check", "list", "format", "status", "heartbeat",
"cron", "lookup", "transform", "filter", "sort", "convert",
}
MODERATE_SIGNALS = {
"write", "generate", "summar", "summarize", "draft", "analyze",
"refactor", "implement", "extract",
}
COMPLEX_SIGNALS = {
"debug", "architect", "design", "security", "adversarial", "ambiguous",
"novel", "debugging", "reason", "why", "migrate", "review",
}
def classify(text: str, extra_keywords: list[str] | None = None) -> str:
"""Classify a task description into a complexity bucket."""
lower = text.lower()
tokens = set(re.findall(r"[a-z0-9_]+", lower))
if extra_keywords:
tokens.update(k.lower() for k in extra_keywords)
if tokens & COMPLEX_SIGNALS:
return COMPLEX
if tokens & ROUTINE_SIGNALS and not tokens & MODERATE_SIGNALS:
return ROUTINE
if tokens & MODERATE_SIGNALS:
return MODERATE
# fallback: no strong signals -> call it routine, keep it cheap
return ROUTINE
def tier_of(bucket: str) -> str:
"""Map a complexity bucket to a tier label."""
return {
ROUTINE: "Tier 1 (cheap)",
MODERATE: "Tier 2 (mid)",
COMPLEX: "Tier 3 (premium)",
}[bucket]
def main() -> None:
if len(sys.argv) < 2:
raise SystemExit("usage: classify_task.py <task> [keyword...]")
task = sys.argv[1]
extra = sys.argv[2:]
bucket = classify(task, extra)
print(f"task: {task}")
print(f"bucket: {bucket}")
print(f"recommend {tier_of(bucket)}")
if __name__ == "__main__":
main()
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |