AGENT SKILL

eval-harness

暂无简介

License 许可
MIT
最近更新
2026-11-01
Stars
★ 3
Forks
1

03 SKILL.md 正文

Eval Harness Skill

A formal evaluation framework for Claude Code sessions, implementing eval-driven development (EDD) principles.

Philosophy

Eval-Driven Development treats evals as the "unit tests of AI development":

  • Define expected behavior BEFORE implementation
  • Run evals continuously during development
  • Track regressions with each change
  • Use pass@k metrics for reliability measurement

Eval Types

Capability Evals

Test if Claude can do something it couldn't before:

[CAPABILITY EVAL: feature-name]
Task: Description of what Claude should accomplish
Success Criteria:
  - [ ] Criterion 1
  - [ ] Criterion 2
  - [ ] Criterion 3
Expected Output: Description of expected result

Regression Evals

Ensure changes don't break existing functionality:

[REGRESSION EVAL: feature-name]
Baseline: SHA or checkpoint name
Tests:
  - existing-test-1: PASS/FAIL
  - existing-test-2: PASS/FAIL
  - existing-test-3: PASS/FAIL
Result: X/Y passed (previously Y/Y)

Grader Types

1. Code-Based Grader

Deterministic checks using code:

# Check if file contains expected pattern
grep -q "export function handleAuth" src/auth.ts && echo "PASS" || echo "FAIL"

# Check if tests pass
npm test -- --testPathPattern="auth" && echo "PASS" || echo "FAIL"

# Check if build succeeds
npm run build && echo "PASS" || echo "FAIL"

2. Model-Based Grader

Use Claude to evaluate open-ended outputs:

[MODEL GRADER PROMPT]
Evaluate the following code change:
1. Does it solve the stated problem?
2. Is it well-structured?
3. Are edge cases handled?
4. Is error handling appropriate?

Score: 1-5 (1=poor, 5=excellent)
Reasoning: [explanation]

3. Human Grader

Flag for manual review:

[HUMAN REVIEW REQUIRED]
Change: Description of what changed
Reason: Why human review is needed
Risk Level: LOW/MEDIUM/HIGH

Metrics

pass@k

"At least one success in k attempts"

  • pass@1: First attempt success rate
  • pass@3: Success within 3 attempts
  • Typical target: pass@3 > 90%

pass^k

"All k trials succeed"

  • Higher bar for reliability
  • pass^3: 3 consecutive successes
  • Use for critical paths

Eval Workflow

1. Define (Before Coding)

## EVAL DEFINITION: feature-xyz

### Capability Evals
1. Can create new user account
2. Can validate email format
3. Can hash password securely

### Regression Evals
1. Existing login still works
2. Session management unchanged
3. Logout flow intact

### Success Metrics
- pass@3 > 90% for capability evals
- pass^3 = 100% for regression evals

2. Implement

Write code to pass the defined evals.

3. Evaluate

# Run capability evals
[Run each capability eval, record PASS/FAIL]

# Run regression evals
npm test -- --testPathPattern="existing"

# Generate report

4. Report

EVAL REPORT: feature-xyz
========================

Capability Evals:
  create-user:     PASS (pass@1)
  validate-email:  PASS (pass@2)
  hash-password:   PASS (pass@1)
  Overall:         3/3 passed

Regression Evals:
  login-flow:      PASS
  session-mgmt:    PASS
  logout-flow:     PASS
  Overall:         3/3 passed

Metrics:
  pass@1: 67% (2/3)
  pass@3: 100% (3/3)

Status: READY FOR REVIEW

Integration Patterns

Pre-Implementation

/eval define feature-name

Creates eval definition file at .claude/evals/feature-name.md

During Implementation

/eval check feature-name

Runs current evals and reports status

Post-Implementation

/eval report feature-name

Generates full eval report

Eval Storage

Store evals in project:

.claude/
  evals/
    feature-xyz.md      # Eval definition
    feature-xyz.log     # Eval run history
    baseline.json       # Regression baselines

Best Practices

  1. Define evals BEFORE coding - Forces clear thinking about success criteria
  2. Run evals frequently - Catch regressions early
  3. Track pass@k over time - Monitor reliability trends
  4. Use code graders when possible - Deterministic > probabilistic
  5. Human review for security - Never fully automate security checks
  6. Keep evals fast - Slow evals don't get run
  7. Version evals with code - Evals are first-class artifacts

Example: Adding Authentication

## EVAL: add-authentication

### Phase 1: Define (10 min)
Capability Evals:
- [ ] User can register with email/password
- [ ] User can login with valid credentials
- [ ] Invalid credentials rejected with proper error
- [ ] Sessions persist across page reloads
- [ ] Logout clears session

Regression Evals:
- [ ] Public routes still accessible
- [ ] API responses unchanged
- [ ] Database schema compatible

### Phase 2: Implement (varies)
[Write code]

### Phase 3: Evaluate
Run: /eval check add-authentication

### Phase 4: Report
EVAL REPORT: add-authentication
==============================
Capability: 5/5 passed (pass@3: 100%)
Regression: 3/3 passed (pass^3: 100%)
Status: SHIP IT

04 同仓库更多 Skills

shengdabai
shengdabai/Tony-Claude-Code-Skills

seedance-prompt-en

Write effective prompts for Jimeng Seedance 2.0 multimodal AI video generation. Use when users want to create video prompts using text, images, videos, and audio inputs with the @ reference system. Covers camera movements, effects replication, video extension, editing, music beat-matching, e-commerce ads, short dramas, and educational content.

★ 3
shengdabai
shengdabai/Tony-Claude-Code-Skills

onepage-pdf

Convert an HTML page into a single continuous-page PDF (one tall page, no pagination breaks), preserving desktop layout, backgrounds and selectable text. Use when the user asks to turn an HTML file/report/proposal into a "single-page PDF", "long PDF", "不分页 PDF", "单页 PDF", "长图式 PDF", or complains that an HTML-to-PDF export breaks into pages or cuts content. Supports optional string redaction with leak verification.

★ 3
shengdabai
shengdabai/Tony-Claude-Code-Skills

native-feel-cross-platform-desktop

Use when the user is designing, prototyping, or rewriting a desktop app that must run on multiple OSes (macOS + Windows, optionally Linux) AND feel indistinguishable from a native app to its users — fast launch, native windowing, native input handling, native materials. Trigger words include "cross-platform desktop", "Electron alternative", "Tauri vs native", "WebView wrapper", "near-native performance", "Raycast architecture", "WebKit/WebView2 quirks", "WKWebView", "system tray app", "global hotkey app", "launcher app". Do NOT trigger this skill for pure web apps, pure mobile apps, or for greenfield projects that have no native-feel requirement.

★ 3
shengdabai
shengdabai/Tony-Claude-Code-Skills

gstack

Router for the gstack skill suite. (gstack)

★ 3
shengdabai
shengdabai/Tony-Claude-Code-Skills

fireworks-tech-graph

Use when the user wants to create any technical diagram - architecture, data flow, flowchart, sequence, agent/memory, or concept map - and export as SVG+PNG. Trigger on: "画图" "帮我画" "生成图" "做个图" "架构图" "流程图" "可视化一下" "出图" "generate diagram" "draw diagram" "visualize" or any system/flow description the user wants illustrated.

★ 3
shengdabai
shengdabai/Tony-Claude-Code-Skills

dbs-video-extract

接收抖音、小红书、微信视频号等短视频链接或分享文案,查询作品/账号数据并提取语音文字稿,按作者和标题归档为 Markdown。用户希望完整解析短视频、取得数据或文字稿时使用;运行前检查 TikHub 与轻抖使用凭证,缺少时用新手能理解的方式引导购买和安全配置。

★ 3