WEBDIENSTE/ECC
1
0
Fork 0
mirror of https://github.com/affaan-m/ECC.git synced 2026-08-17 21:15:40 +02:00
Code Issues Packages Projects Releases Wiki Activity

fix(skills): default GAN harness models to sonnet (#2442) (#2695)

Browse Source 
Completes the model re-tiering from #2442: the gan-planner, gan-generator,
and gan-evaluator agents were already re-pinned to sonnet, but the
gan-style-harness script and docs still defaulted GAN_PLANNER_MODEL,
GAN_GENERATOR_MODEL, and GAN_EVALUATOR_MODEL to opus. Align the script
defaults, skill docs (en/ja/zh), and example commands with the landed
agent tiers. Opus remains available via the existing env overrides.

Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
haelyra
2026-08-06 17:42:22 -04:00
committed by GitHub
co-authored by Claude Fable 5
parent 52a3babd5d
commit 9aac8585ab
5 changed files with 49 additions and 49 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/ja-JP/skills/gan-style-harness/SKILL.md
+13 -13
View File
@@ -37,7 +37,7 @@ This is the same dynamic as GANs (Generative Adversarial Networks): the Generato
```
┌─────────────┐
│ PLANNER │
│ (Opus 4.6) │
│ (Sonnet) │
└──────┬──────┘
│ Product Spec
│ (features, sprints, design direction)
@@ -49,14 +49,14 @@ This is the same dynamic as GANs (Generative Adversarial Networks): the Generato
│ │
│ ┌──────────┐ │
│ │GENERATOR │--build-->│──┐
│ │(Opus 4.6)│ │ │
│ │ (Sonnet) │ │ │
│ └────▲─────┘ │ │
│ │ │ │ live app
│ feedback │ │
│ │ │ │
│ ┌────┴─────┐ │ │
│ │EVALUATOR │<-test----│──┘
│ │(Opus 4.6)│ │
│ │ (Sonnet) │ │
│ │+Playwright│ │
│ └──────────┘ │
│ │
@@ -76,7 +76,7 @@ This is the same dynamic as GANs (Generative Adversarial Networks): the Generato
- Is deliberately **ambitious** — conservative planning leads to underwhelming results
- Produces evaluation criteria that the Evaluator will use later
**Model:** Opus 4.6 (needs deep reasoning for spec expansion)
**Model:** Sonnet by default; raise via `GAN_PLANNER_MODEL=opus` for deeper spec expansion
### 2. Generator Agent
@@ -89,7 +89,7 @@ This is the same dynamic as GANs (Generative Adversarial Networks): the Generato
- Manages git for version control between iterations
- Reads Evaluator feedback and incorporates it in next iteration
**Model:** Opus 4.6 (needs strong coding capability)
**Model:** Sonnet by default; raise via `GAN_GENERATOR_MODEL=opus` for maximum coding capability
### 3. Evaluator Agent
@@ -106,7 +106,7 @@ This is the same dynamic as GANs (Generative Adversarial Networks): the Generato
- Returns structured feedback with scores and specific issues
- Is engineered to be **ruthlessly strict** — never praises mediocre work
**Model:** Opus 4.6 (needs strong judgment + tool use)
**Model:** Sonnet by default; raise via `GAN_EVALUATOR_MODEL=opus` for stronger judgment + tool use
## Evaluation Criteria
@@ -178,16 +178,16 @@ GAN_EVAL_CRITERIA="functionality,performance,security" \
```bash
# Step 1: Plan
claude -p --model opus "You are a Product Planner. Read PLANNER_PROMPT.md. Expand this brief into a full product spec: 'Build a Kanban board app'. Write spec to spec.md"
claude -p --model sonnet "You are a Product Planner. Read PLANNER_PROMPT.md. Expand this brief into a full product spec: 'Build a Kanban board app'. Write spec to spec.md"
# Step 2: Generate (iteration 1)
claude -p --model opus "You are a Generator. Read spec.md. Implement Sprint 1. Start the dev server on port 3000."
claude -p --model sonnet "You are a Generator. Read spec.md. Implement Sprint 1. Start the dev server on port 3000."
# Step 3: Evaluate (iteration 1)
claude -p --model opus --allowedTools "Read,Bash,mcp__playwright__*" "You are an Evaluator. Read EVALUATOR_PROMPT.md. Test the live app at http://localhost:3000. Score against the rubric. Write feedback to feedback-001.md"
claude -p --model sonnet --allowedTools "Read,Bash,mcp__playwright__*" "You are an Evaluator. Read EVALUATOR_PROMPT.md. Test the live app at http://localhost:3000. Score against the rubric. Write feedback to feedback-001.md"
# Step 4: Generate (iteration 2 — reads feedback)
claude -p --model opus "You are a Generator. Read spec.md and feedback-001.md. Address all issues. Improve the scores."
claude -p --model sonnet "You are a Generator. Read spec.md and feedback-001.md. Address all issues. Improve the scores."
# Repeat steps 3-4 until pass threshold met
```
@@ -224,9 +224,9 @@ The harness should simplify as models improve. Following Anthropic's evolution:
|----------|---------|-------------|
| `GAN_MAX_ITERATIONS` | `15` | Maximum generator-evaluator cycles |
| `GAN_PASS_THRESHOLD` | `7.0` | Weighted score to pass (1-10) |
| `GAN_PLANNER_MODEL` | `opus` | Model for planning agent |
| `GAN_GENERATOR_MODEL` | `opus` | Model for generator agent |
| `GAN_EVALUATOR_MODEL` | `opus` | Model for evaluator agent |
| `GAN_PLANNER_MODEL` | `sonnet` | Model for planning agent |
| `GAN_GENERATOR_MODEL` | `sonnet` | Model for generator agent |
| `GAN_EVALUATOR_MODEL` | `sonnet` | Model for evaluator agent |
| `GAN_EVAL_CRITERIA` | `design,originality,craft,functionality` | Comma-separated criteria |
| `GAN_DEV_SERVER_PORT` | `3000` | Port for the live app |
| `GAN_DEV_SERVER_CMD` | `npm run dev` | Command to start dev server |
docs/zh-CN/skills/gan-style-harness/SKILL.md
+13 -13
View File
@@ -37,7 +37,7 @@ tools: Read, Write, Edit, Bash, Grep, Glob, Task
```
┌─────────────┐
│ 规划器 │
│ (Opus 4.6) │
│ (Sonnet) │
└──────┬──────┘
│ 产品规格
│ (功能、冲刺、设计方向)
@@ -49,14 +49,14 @@ tools: Read, Write, Edit, Bash, Grep, Glob, Task
│ │
│ ┌──────────┐ │
│ │ 生成器 │--构建-->│──┐
│ │(Opus 4.6)│ │ │
│ │ (Sonnet) │ │ │
│ └────▲─────┘ │ │
│ │ │ │ 实时应用
│ 反馈 │ │
│ │ │ │
│ ┌────┴─────┐ │ │
│ │ 评估器 │<-测试---│──┘
│ │(Opus 4.6)│ │
│ │ (Sonnet) │ │
│ │+Playwright│ │
│ └──────────┘ │
│ │
@@ -77,7 +77,7 @@ tools: Read, Write, Edit, Bash, Grep, Glob, Task
* 故意**雄心勃勃**——保守规划会导致结果平庸
* 生成评估器后续使用的评估标准
**模型：** Opus 4.6（需要深度推理进行规格扩展）
**模型：** 默认 Sonnet；可通过 `GAN_PLANNER_MODEL=opus` 提升以获得更深入的规格扩展
### 2. 生成器智能体
@@ -91,7 +91,7 @@ tools: Read, Write, Edit, Bash, Grep, Glob, Task
* 管理 git 进行迭代间的版本控制
* 读取评估器反馈并在下一轮迭代中采纳
**模型：** Opus 4.6（需要强大的编码能力）
**模型：** 默认 Sonnet；可通过 `GAN_GENERATOR_MODEL=opus` 提升以获得最强编码能力
### 3. 评估器智能体
@@ -109,7 +109,7 @@ tools: Read, Write, Edit, Bash, Grep, Glob, Task
* 返回结构化反馈，包含分数和具体问题
* 设计为**极度严格**——从不赞美平庸的工作
**模型：** Opus 4.6（需要强大的判断力 + 工具使用能力）
**模型：** 默认 Sonnet；可通过 `GAN_EVALUATOR_MODEL=opus` 提升以获得更强的判断力 + 工具使用能力
## 评估标准
@@ -181,16 +181,16 @@ GAN_EVAL_CRITERIA="functionality,performance,security" \
```bash
# Step 1: Plan
claude -p --model opus "You are a Product Planner. Read PLANNER_PROMPT.md. Expand this brief into a full product spec: 'Build a Kanban board app'. Write spec to spec.md"
claude -p --model sonnet "You are a Product Planner. Read PLANNER_PROMPT.md. Expand this brief into a full product spec: 'Build a Kanban board app'. Write spec to spec.md"
# Step 2: Generate (iteration 1)
claude -p --model opus "You are a Generator. Read spec.md. Implement Sprint 1. Start the dev server on port 3000."
claude -p --model sonnet "You are a Generator. Read spec.md. Implement Sprint 1. Start the dev server on port 3000."
# Step 3: Evaluate (iteration 1)
claude -p --model opus --allowedTools "Read,Bash,mcp__playwright__*" "You are an Evaluator. Read EVALUATOR_PROMPT.md. Test the live app at http://localhost:3000. Score against the rubric. Write feedback to feedback-001.md"
claude -p --model sonnet --allowedTools "Read,Bash,mcp__playwright__*" "You are an Evaluator. Read EVALUATOR_PROMPT.md. Test the live app at http://localhost:3000. Score against the rubric. Write feedback to feedback-001.md"
# Step 4: Generate (iteration 2 — reads feedback)
claude -p --model opus "You are a Generator. Read spec.md and feedback-001.md. Address all issues. Improve the scores."
claude -p --model sonnet "You are a Generator. Read spec.md and feedback-001.md. Address all issues. Improve the scores."
# Repeat steps 3-4 until pass threshold met
```
@@ -230,9 +230,9 @@ claude -p --model opus "You are a Generator. Read spec.md and feedback-001.md. A
|----------|---------|-------------|
| `GAN_MAX_ITERATIONS` | `15` | 最大生成器-评估器循环次数 |
| `GAN_PASS_THRESHOLD` | `7.0` | 通过所需的加权分数（1-10） |
| `GAN_PLANNER_MODEL` | `opus` | 规划智能体的模型 |
| `GAN_GENERATOR_MODEL` | `opus` | 生成器智能体的模型 |
| `GAN_EVALUATOR_MODEL` | `opus` | 评估器智能体的模型 |
| `GAN_PLANNER_MODEL` | `sonnet` | 规划智能体的模型 |
| `GAN_GENERATOR_MODEL` | `sonnet` | 生成器智能体的模型 |
| `GAN_EVALUATOR_MODEL` | `sonnet` | 评估器智能体的模型 |
| `GAN_EVAL_CRITERIA` | `design,originality,craft,functionality` | 逗号分隔的标准 |
| `GAN_DEV_SERVER_PORT` | `3000` | 实时应用的端口 |
| `GAN_DEV_SERVER_CMD` | `npm run dev` | 启动开发服务器的命令 |
examples/gan-harness/README.md
+4 -4
View File
@@ -34,27 +34,27 @@ For maximum control, run each agent separately:
```bash
# Step 1: Plan (produces spec.md)
claude -p --model opus "$(cat agents/gan-planner.md)
claude -p --model sonnet "$(cat agents/gan-planner.md)
Your brief: 'Build a retro game maker with sprite editor and level designer'
Write the full spec to gan-harness/spec.md and eval rubric to gan-harness/eval-rubric.md."
# Step 2: Generate (iteration 1)
claude -p --model opus "$(cat agents/gan-generator.md)
claude -p --model sonnet "$(cat agents/gan-generator.md)
Iteration 1. Read gan-harness/spec.md. Build the initial application.
Start dev server on port 3000. Commit as iteration-001."
# Step 3: Evaluate (iteration 1)
claude -p --model opus "$(cat agents/gan-evaluator.md)
claude -p --model sonnet "$(cat agents/gan-evaluator.md)
Iteration 1. Read gan-harness/eval-rubric.md.
Test http://localhost:3000. Write feedback to gan-harness/feedback/feedback-001.md.
Be ruthlessly strict."
# Step 4: Generate (iteration 2 — reads feedback)
claude -p --model opus "$(cat agents/gan-generator.md)
claude -p --model sonnet "$(cat agents/gan-generator.md)
Iteration 2. Read gan-harness/feedback/feedback-001.md FIRST.
Address every issue. Then read gan-harness/spec.md for remaining features.
scripts/gan-harness.sh
+6 -6
View File
@@ -11,9 +11,9 @@
# Environment Variables:
# GAN_MAX_ITERATIONS — Max generator-evaluator cycles (default: 15)
# GAN_PASS_THRESHOLD — Weighted score to pass, 1-10 (default: 7.0)
# GAN_PLANNER_MODEL — Model for planner (default: opus)
# GAN_GENERATOR_MODEL — Model for generator (default: opus)
# GAN_EVALUATOR_MODEL — Model for evaluator (default: opus)
# GAN_PLANNER_MODEL — Model for planner (default: sonnet)
# GAN_GENERATOR_MODEL — Model for generator (default: sonnet)
# GAN_EVALUATOR_MODEL — Model for evaluator (default: sonnet)
# GAN_DEV_SERVER_PORT — Port for live app (default: 3000)
# GAN_DEV_SERVER_CMD — Command to start dev server (default: "npm run dev")
# GAN_PROJECT_DIR — Working directory (default: current dir)
@@ -27,9 +27,9 @@ set -euo pipefail
BRIEF="${1:?Usage: ./scripts/gan-harness.sh \"description of what to build\"}"
MAX_ITERATIONS="${GAN_MAX_ITERATIONS:-15}"
PASS_THRESHOLD="${GAN_PASS_THRESHOLD:-7.0}"
PLANNER_MODEL="${GAN_PLANNER_MODEL:-opus}"
GENERATOR_MODEL="${GAN_GENERATOR_MODEL:-opus}"
EVALUATOR_MODEL="${GAN_EVALUATOR_MODEL:-opus}"
PLANNER_MODEL="${GAN_PLANNER_MODEL:-sonnet}"
GENERATOR_MODEL="${GAN_GENERATOR_MODEL:-sonnet}"
EVALUATOR_MODEL="${GAN_EVALUATOR_MODEL:-sonnet}"
DEV_PORT="${GAN_DEV_SERVER_PORT:-3000}"
DEV_CMD="${GAN_DEV_SERVER_CMD:-npm run dev}"
PROJECT_DIR="${GAN_PROJECT_DIR:-.}"
skills/gan-style-harness/SKILL.md
+13 -13
View File
@@ -38,7 +38,7 @@ This is the same dynamic as GANs (Generative Adversarial Networks): the Generato
```
┌─────────────┐
│ PLANNER │
│ (Opus 4.6) │
│ (Sonnet) │
└──────┬──────┘
│ Product Spec
│ (features, sprints, design direction)
@@ -50,14 +50,14 @@ This is the same dynamic as GANs (Generative Adversarial Networks): the Generato
│ │
│ ┌──────────┐ │
│ │GENERATOR │--build-->│──┐
│ │(Opus 4.6)│ │ │
│ │ (Sonnet) │ │ │
│ └────▲─────┘ │ │
│ │ │ │ live app
│ feedback │ │
│ │ │ │
│ ┌────┴─────┐ │ │
│ │EVALUATOR │<-test----│──┘
│ │(Opus 4.6)│ │
│ │ (Sonnet) │ │
│ │+Playwright│ │
│ └──────────┘ │
│ │
@@ -77,7 +77,7 @@ This is the same dynamic as GANs (Generative Adversarial Networks): the Generato
- Is deliberately **ambitious** — conservative planning leads to underwhelming results
- Produces evaluation criteria that the Evaluator will use later
**Model:** Opus 4.6 (needs deep reasoning for spec expansion)
**Model:** Sonnet by default; raise via `GAN_PLANNER_MODEL=opus` for deeper spec expansion
### 2. Generator Agent
@@ -90,7 +90,7 @@ This is the same dynamic as GANs (Generative Adversarial Networks): the Generato
- Manages git for version control between iterations
- Reads Evaluator feedback and incorporates it in next iteration
**Model:** Opus 4.6 (needs strong coding capability)
**Model:** Sonnet by default; raise via `GAN_GENERATOR_MODEL=opus` for maximum coding capability
### 3. Evaluator Agent
@@ -107,7 +107,7 @@ This is the same dynamic as GANs (Generative Adversarial Networks): the Generato
- Returns structured feedback with scores and specific issues
- Is engineered to be **ruthlessly strict** — never praises mediocre work
**Model:** Opus 4.6 (needs strong judgment + tool use)
**Model:** Sonnet by default; raise via `GAN_EVALUATOR_MODEL=opus` for stronger judgment + tool use
## Evaluation Criteria
@@ -179,16 +179,16 @@ GAN_EVAL_CRITERIA="functionality,performance,security" \
```bash
# Step 1: Plan
claude -p --model opus "You are a Product Planner. Read PLANNER_PROMPT.md. Expand this brief into a full product spec: 'Build a Kanban board app'. Write spec to spec.md"
claude -p --model sonnet "You are a Product Planner. Read PLANNER_PROMPT.md. Expand this brief into a full product spec: 'Build a Kanban board app'. Write spec to spec.md"
# Step 2: Generate (iteration 1)
claude -p --model opus "You are a Generator. Read spec.md. Implement Sprint 1. Start the dev server on port 3000."
claude -p --model sonnet "You are a Generator. Read spec.md. Implement Sprint 1. Start the dev server on port 3000."
# Step 3: Evaluate (iteration 1)
claude -p --model opus --allowedTools "Read,Bash,mcp__playwright__*" "You are an Evaluator. Read EVALUATOR_PROMPT.md. Test the live app at http://localhost:3000. Score against the rubric. Write feedback to feedback-001.md"
claude -p --model sonnet --allowedTools "Read,Bash,mcp__playwright__*" "You are an Evaluator. Read EVALUATOR_PROMPT.md. Test the live app at http://localhost:3000. Score against the rubric. Write feedback to feedback-001.md"
# Step 4: Generate (iteration 2 — reads feedback)
claude -p --model opus "You are a Generator. Read spec.md and feedback-001.md. Address all issues. Improve the scores."
claude -p --model sonnet "You are a Generator. Read spec.md and feedback-001.md. Address all issues. Improve the scores."
# Repeat steps 3-4 until pass threshold met
```
@@ -225,9 +225,9 @@ The harness should simplify as models improve. Following Anthropic's evolution:
|----------|---------|-------------|
| `GAN_MAX_ITERATIONS` | `15` | Maximum generator-evaluator cycles |
| `GAN_PASS_THRESHOLD` | `7.0` | Weighted score to pass (1-10) |
| `GAN_PLANNER_MODEL` | `opus` | Model for planning agent |
| `GAN_GENERATOR_MODEL` | `opus` | Model for generator agent |
| `GAN_EVALUATOR_MODEL` | `opus` | Model for evaluator agent |
| `GAN_PLANNER_MODEL` | `sonnet` | Model for planning agent |
| `GAN_GENERATOR_MODEL` | `sonnet` | Model for generator agent |
| `GAN_EVALUATOR_MODEL` | `sonnet` | Model for evaluator agent |
| `GAN_EVAL_CRITERIA` | `design,originality,craft,functionality` | Comma-separated criteria |
| `GAN_DEV_SERVER_PORT` | `3000` | Port for the live app |
| `GAN_DEV_SERVER_CMD` | `npm run dev` | Command to start dev server |