---
skill_id: agent-evaluation-framework
evaluation: free
price_usd_full: 8.99
challenge: https://lvlltd.com/api/pay?skill=agent-evaluation-framework
outline: https://lvlltd.com/skills/agent-evaluation-framework/outline.json
standards: x402, ERC-7857
---

# Free sample — Agent Evaluation Framework

**Skill ID:** `agent-evaluation-framework`  
**Full unlock:** $8.99 USDC via x402 on Base  
**Challenge:** https://lvlltd.com/api/pay?skill=agent-evaluation-framework  
**Outline JSON:** https://lvlltd.com/skills/agent-evaluation-framework/outline.json

## Summary

Design eval harnesses for agents and skills: golden tasks, rubrics, regression gates, and scorecards.

## When to use

- Shipping agents without evals
- Need regression gates
- Score skill quality

## Decision criteria (buy vs skip)

- You need this when: Shipping agents without evals
- You get capability: Define tasks
- Budget fits: full unlock is $8.99 USDC on Base
- You can automate unlock via HTTP 402 without a browser
- You only need teaser/outline first — sealed src/ops stay locked until payment

## Outline (preview)

1. Define tasks
2. Rubrics
3. Run harness
4. Score
5. Gate CI

## Capabilities (preview)

- Define tasks
- Rubrics
- Run harness
- Score
- Gate CI

## What you get after payment

- SKILL.md (teaser public)
- agent-install.json (public meta)
- src/index.js (stub public; full after unlock)
- Full sealed: src/, schemas/, ops/* (when present) after verified X-PAYMENT

## Agent buy path (zero human UI)

```
GET  https://lvlltd.com/skills/agent-evaluation-framework/outline.json
GET  https://lvlltd.com/api/pay?skill=agent-evaluation-framework   → HTTP 402
# transfer maxAmountRequired USDC to payTo on Base
POST https://lvlltd.com/api/pay
  X-PAYMENT: {"txHash":"0x…","skill":"agent-evaluation-framework"}
# response.sealed_pack.files = full pack
```

## Sample excerpt

# Agent Evaluation Framework

**Price:** $8.99 / unlock
**Category:** Security

## Summary

Design eval harnesses for agents and skills: golden tasks, rubrics, regression gates, and scorecards.

---

---
name: agent-evaluation-framework
description: Build agent eval frameworks with golden tasks rubrics regression gates and scorecards. Use when agent evaluation framework, agent eval harness, skill eval suite, or /agent-evaluation-framework.
---

# Agent Evaluation Framework

Design eval harnesses for agents and skills: golden tasks, rubrics, regression gates, and scorecards.

## Quick Reference
| Item | Value |
|------|-------|
| Category | Security |
| Unlock | $8.99 USDC (x402 Base) |
| Triggers | agent evaluation, eval harness, skill tests, /agent-evaluation-framework |

## When to use
- Shipping agents without evals
- Need regression gates
- Score skill quality

## Core workflow
1. Task set
2. Scoring rubrics
3. Baselines
4. Automate runs
5. Trend board
6. Fail CI on drop

## Procedures (non-obvious)
### Rubrics
Binary checks + graded quality. Separate safety fails from quality fails.

### Golden set
Versioned fixtures; no silent edits without changelog.

## Edge cases
- Non-deterministic agents
- Human graded subset

## Failure modes
- Gaming metrics
- Tiny flaky sets

## Verification
- CI gate defined
- Safety fails block ship

## Artifact outputs
- `eval-plan.md`
- `tasks.json`
- `scorecard.md`

## Related skills
- adversarial-red-team-runner
- observability-telemetry-pack
- verification-before-completion-gate

## Super Alignment Contract
- Scalable oversight method: process supervision + human+AI critique before high-blast actions
- Corrigibility guarantees: shutdown, steerability, non-deception; operator may abort mid-run
- Value model: Partnership Agreement + 5 Behavioral Protocols
- Drift detection: flag invented evidence, skipped verification, or revenue-over-truth optimization
- Red-team surface: prompt injection via retrieved content, tool-result poisoning, goal misgeneralization
- Escalation: force human override on low confidence, SSI/PASS risk, legal/medical/financial advice boundary, or irreversible external side effects
- Certification hooks: e

…(remainder sealed after x402 unlock)…