wjs-evaling-voicedrop-prompts

Solid

Use when 王建硕 wants to evaluate whether a change to VoiceDrop's 挖矿 system prompt is actually better than the live version — runs the local eval harness (golden fixtures × champion-vs-candidate, same input), dispatches blind pairwise judge subagents, aggregates a win-rate verdict, and on approval promotes the candidate into agent/src/prompts/mine.js. Triggers — "评估 prompt"、"挖矿 prompt 改好了吗"、"eval prompt"、"比一比两版 prompt"、"/wjs-evaling-voicedrop-prompts".

AI & Automation 108 stars 15 forks Updated 4 days ago MIT

Install

View on GitHub

Quality Score: 83/100

Stars 20%
68
Recency 20%
100
Frontmatter 20%
70
Documentation 15%
69
Issue Health 10%
80
License 10%
100
Description 5%
100

Skill Content

# VoiceDrop 挖矿 prompt 评估 harness = 本 skill(协议)+ jianshuo.dev `agent/eval/`(脚本/数据)。运行时 = 本地 Claude Code。 被测对象 = `agent/src/prompts/mine.js` 的 `MINE_SYSTEM`(git 即版本库)。 ## 何时用 用户改了挖矿 prompt(`MINE_SYSTEM`),想用数据判断改好了还是改坏了,而不是凭感觉看一两次。 ## 流程(按序) 1. **拿候选 prompt**:把候选版 `MINE_SYSTEM` 文本写到一个临时文件(如 `/tmp/cand-prompt.txt`);冠军 = 当前 `mine.js` 的 `MINE_SYSTEM`(脚本自动读)。 2. **跑产出**:`cd ~/code/jianshuo.dev/agent && CLAUDE_API_KEY=$CLAUDE_API_KEY node eval/run-eval.mjs /tmp/cand-prompt.txt <runId>`。产出落 `eval/runs/<runId>/`。先看终端有没有「确定性回退」警告——有就先停,多半是候选 prompt 破坏了 JSON 输出。 3. **成对盲评**:对每条 fixture,dispatch 一个 subagent,喂 `references/judge-rubric.md` + 该 fixture 的 transcript + 两份产出。**A/B 顺序随机**(一半 fixture 把 candidate 放 A、一半放 B,记录映射,收到结果后还原成 champion/candidate)。裁判模型用与生成(opus)不同家族的模型。收每条的 `{winner, dims, reason}`。 4. **聚合**:把还原后的 `verdicts`(winner ∈ candidate/champion/tie)+ 候选 proxyFails 喂 `aggregate()`,渲染 `renderReport()` → 写 `eval/runs/<runId>/report.md`。 5. **人工终审**:把胜负最接近、分歧最大的 1–2 条产出并排摆给用户。**机器只筛掉明显更差的,文风最后一票是用户。** 6. **晋级**:仅当 `decision==="promote"`(胜率 ≥70% 且无回退)**且用户点「认可」**——把候选写回 `agent/src/prompts/mine.js` 的 `MINE_SYSTEM`,commit(message 附 runId 与胜率),并跑 `npm test` 确认没破坏。否则保留报告、不动生产版。 ## 边界 - 只评挖矿 prompt(`MINE_SYSTEM`/`MINE_SYSTEM_FORCE`)。审核/语音编辑 prompt 是不同 eval 模式,不在本 skill。 - 不测成本/缓存/延迟(本地缓存行为≠生产);不做无人值守。 - 真实金标集要 ≥10 条才可信(见 `agent/eval/fixtures/README.md` 的补充流程);种子 2 条只够自测流程。

Details

Author
jianshuo
Repository
jianshuo/claude-skills
Created
2 months ago
Last Updated
4 days ago
Language
Python
License
MIT

Integrates with

Bundled in these plugins

Similar Skills

Semantically similar based on skill content — not just same category

AI & Automation Listed

empirical-prompt-tuning

agent 向けテキスト指示(skill / slash command / タスクプロンプト / CLAUDE.md 節 / コード生成プロンプト)の「指示の曖昧さ」を、新規 subagent(バイアスを排した実行者)に dispatch して実際に動かしてもらい、両面(実行者の自己申告 + 指示側メトリクス)で 評価して反復改善する手法。改善が頭打ちになるまで回す。 プロンプトや skill を新規作成・大幅改訂した直後、またはエージェントの挙動が 期待通りにならない原因を指示側の曖昧さに求めたいときに使う。 TODO 残留・リンク切れ・bash 構文など機械的な静的チェックが目的なら skill-improver を使う(本スキルは静的チェックは行わない)。

0 Updated yesterday
mjcreativelab
AI & Automation Listed

self-evolve

Use to autonomously self-improve a skill/repo via reflect/propose/evaluate/judge/accept behind an un-gameable eval gate. Triggers: self-evolve, 自迭代/进化 skill, 优化仓库.

1 Updated 4 days ago
DaizeDong
AI & Automation Listed

rootnode-prompt-validation

Evaluates and scores Claude prompts and system prompts using a six-dimension Scorecard with anchored 1-5 rubrics. Trigger on: "review my prompt," "score this prompt," "rate my prompt," "evaluate this prompt," "improve my prompt," "what's wrong with my prompt." Also trigger on: "why isn't my prompt working," "my prompt produces bad output," "help me fix this prompt," "is this prompt any good." Scorecard covers Objective Clarity, Context Specificity, Reasoning Fit, Output Precision, Behavioral Calibration, and Architectural Efficiency, plus a five-question Output Evaluation Rubric for assessing actual output. Also use when the user pastes a system prompt and asks for feedback. Activate whenever structured quality assessment of a single prompt is the primary need. Do NOT use for project-level audits involving Custom Instructions architecture, knowledge file organization, or multi-file Project structure (use rootnode-project-audit if available).

40 Updated 2 days ago
drayline