improvement-discriminator当需要对改进候选多人盲审打分、用 LLM 做语义评估、判断候选是否应被接受、或打分结果全是 hold 想知道为什么时使用。支持 --panel 多审阅者盲审和 --llm-judge 语义评估。不用于结构评估(用 improvement-learner)或门禁决策(用 improvement-gate)。
Install via ClawdBot CLI:
clawdbot install lanyasheng/improvement-discriminatorGrade Fair — based on market validation, documentation quality, package completeness, maintenance status, and authenticity signals.
Potentially destructive shell commands in tool definitions
eval(Audited Apr 16, 2026 · audit v1.0
Generated Apr 30, 2026
一个团队需要对多个改进候选进行盲审打分,以降低单人偏见。使用 --panel 和 --llm-judge 组合,每个审阅者独立打分并生成认知标签(CONSENSUS/VERIFIED/DISPUTED),最后输出排名供 executor 按优先级执行。
产品经理需要对改进候选进行语义评估(清晰度、具体性、一致性、安全性),使用 --llm-judge 模式。LLM 对每个维度打分并给出决策(accept/conditional/reject),提供可解释的评分明细,辅助判断候选是否应被接受。
当所有候选都被标为hold时,通过分析 risk_penalty 过高或缺少 source_refs 等常见原因进行调试。使用 --verbose 查看详细评分过程,找出问题根源。
在大型项目中,单一审阅者可能因个人偏好系统性高/低估某些候选。通过启用 --panel 盲审,多个审阅者独立评分,使用认知标签衡量分歧,从而减少偏差,提高评分公正性。
团队需要追踪多轮改进中候选质量的趋势。使用 discriminator 对每一轮的候选独立打分(--llm-judge 或 --panel),比较各维度得分变化,评估改进效果。
提供 API,让用户上传改进候选,返回带有多维度评分和认知标签的结果。按 API 调用次数或 token 消耗收费。
结合 generator、discriminator、gate、executor 等技能,提供完整的改进候选生命周期管理 SaaS。企业按候选数量或用户数付费。
为客户定制 discriminator 的启发式规则、LLM 评估维度(如行业特定安全性要求),收取一次性项目费加后期维护费。
💬 Integration Tip
确保 generator 产出标准格式的 candidates.json,再调用 discriminator;可通过 --verbose 调试评分逻辑。
Scored Apr 19, 2026
Remove signs of AI-generated writing from text. Use when editing or reviewing text to make it sound more natural and human-written. Based on Wikipedia's comprehensive "Signs of AI writing" guide. Detects and fixes patterns including: inflated symbolism, promotional language, superficial -ing analyses, vague attributions, em dash overuse, rule of three, AI vocabulary words, negative parallelisms, and excessive conjunctive phrases.
Humanize AI-generated text to bypass detection. This humanizer rewrites ChatGPT, Claude, and GPT content to sound natural and pass AI detectors like GPTZero,...
Humanize AI-generated text by detecting and removing patterns typical of LLM output. Rewrites text to sound natural, specific, and human. Uses 24 pattern detectors, 500+ AI vocabulary terms across 3 tiers, and statistical analysis (burstiness, type-token ratio, readability) for comprehensive detection. Use when asked to humanize text, de-AI writing, make content sound more natural/human, review writing for AI patterns, score text for AI detection, or improve AI-generated drafts. Covers content, language, style, communication, and filler categories.
去除文本中的 AI 生成痕迹。适用于编辑或审阅文本,使其听起来更自然、更像人类书写。 基于维基百科的"AI 写作特征"综合指南。检测并修复以下模式:夸大的象征意义、 宣传性语言、以 -ing 结尾的肤浅分析、模糊的归因、破折号过度使用、三段式法则、 AI 词汇、否定式排比、过多的连接性短语。
Collaborative thinking partner for exploring complex problems through questioning
You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation.