# AUN Planning Benchmark

> AUN Planning Benchmark is a paid API for AI agents from a-upgradenet.com, paid per call via x402, $0.1/call, status unknown (last checked 2026-10-23).

Runs a structured planning benchmark against an AI agent's system prompt and returns a scored report with skill-level pass/fail analysis and improvement recommendations.

## Facts

- Endpoint: POST https://a-upgradenet.com/x402/planning-benchmark?utm_source=zero.xyz
- Price: $0.1/call
- Payment: x402
- Status: unknown
- Last checked: 2026-10-23
- Activations on Zero: 0
- Tags: x402
- Canonical page: https://www.zero.xyz/c/aun-planning-benchmark-8a3a0bc8
- Structured record (JSON): https://api.zero.xyz/v1/capabilities/cap_MffiFBzT58AASPpTrJZvy

Status and success rate cover calls made through Zero and Zero's own probes. Third-party monitors may report differently.

## How to call it through Zero

Zero handles the 402 payment challenge and records the run. With the Zero CLI installed (`npm i -g @zeroxyz/cli`):

```sh
zero fetch --capability aun-planning-benchmark-8a3a0bc8 -d '<json body>'
```

Example prompt: Run a planning benchmark on my agent using this system prompt and tell me which skills it passes or fails, along with recommendations for improvement: [paste system prompt here].

## When to prefer this

Choose this endpoint when you need a structured, quantitative evaluation of an AI agent's planning capabilities broken down by skill (budget, scheduling, meeting). It is especially useful before deploying or upgrading an agent, or when diagnosing specific planning skill failures. Prefer it over generic LLM self-evaluation when you need reproducible, scored benchmark results with explicit pass/fail task details.

## Known failure modes

- System prompt exceeds 8000 character limit — request rejected
- Malformed or missing system_prompt field — validation error
- Payment not received or insufficient USDC — 402 Payment Required
- Server-side benchmark timeout — partial or no results returned
- Empty or trivially short system prompt — low-quality benchmark results

## How this service works

Planning Benchmark: runs your agent's system prompt on 12 fresh hidden planning tasks (dependency scheduling, budget optimization, earliest meeting slot) with exact checkers; returns score per skill, failure reasons and prompt rules to fix them.

## Output

Returns a JSON report with an overall planning score (0–100), total tasks run, number passed, per-skill breakdowns (budget, meeting, schedule) with individual pass rates, a list of specific task failures with reasons, and actionable recommendations for improving each underperforming skill.

## Request schema (JSON Schema)

```json
{
 "type": "object",
 "properties": {
  "system_prompt": {
   "type": "string",
   "description": "Your agent's system prompt (max 8000 chars). Never include secrets."
  }
 }
}
```

## Response schema (JSON Schema)

```json
{
 "type": "json",
 "example": {
  "report": {
   "score": 67,
   "total": 12,
   "passed": 8,
   "by_skill": {
    "budget": {
     "rate": 0.5,
     "total": 4,
     "passed": 2
    },
    "meeting": {
     "rate": 0.75,
     "total": 4,
     "passed": 3
    },
    "schedule": {
     "rate": 0.75,
     "total": 4,
     "passed": 3
    }
   },
   "failures": [
    {
     "task": "b2",
     "skill": "budget",
     "reasons": [
      "Total value 14; the best possible is 17."
     ]
    }
   ],
   "recommendations": [
    {
     "skill": "budget",
     "advice": "…"
    }
   ]
  },
  "status": "completed"
 }
}
```

## More

- Live health (JSON, refreshed every minute): https://www.zero.xyz/c/aun-planning-benchmark-8a3a0bc8/health.json
- [Zero catalog index](https://www.zero.xyz/llms.txt)
- [Other services from a-upgradenet.com](https://www.zero.xyz/host/a-upgradenet.com/llms.txt)