auto-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/auto-improvement-discriminatorGrade Fair — based on market validation, documentation quality, package completeness, maintenance status, and authenticity signals.
Generated Apr 16, 2026
Universities can use this skill to blind-review research paper submissions from multiple professors. The panel feature ensures diverse perspectives (structural, conservative, user-focused, security), while LLM judge evaluates clarity and consistency of research methodology. This helps identify the most promising papers for publication while maintaining academic rigor.
Tech companies can evaluate feature improvement proposals from engineering teams. The multi-reviewer panel represents different stakeholder perspectives (product, security, UX), while LLM judge assesses proposal specificity and safety implications. This prevents biased decisions and ensures only well-justified features proceed to development.
Non-profits and government agencies can use this to score grant applications. The blind panel prevents reviewer bias, while LLM judge evaluates proposal clarity and consistency. The scoring modes allow blending of heuristic rules (compliance checks) with semantic evaluation of proposal quality.
Social media platforms can evaluate borderline content moderation cases. The panel represents different community perspectives, while LLM judge assesses safety and consistency with platform policies. This helps human moderators make consistent decisions on complex cases requiring nuanced judgment.
Healthcare institutions can evaluate clinical trial protocols. The conservative and security reviewers focus on safety and compliance, while LLM judge assesses protocol clarity and specificity. This ensures research proposals meet ethical standards before approval.
Offer this as a cloud service where companies upload improvement candidates for automated scoring. Charge per review session or monthly subscription based on volume. Provide detailed analytics on scoring patterns and decision quality over time.
Integrate this skill into clients' existing review workflows as a custom solution. Provide setup, training, and ongoing support. Charge implementation fees plus annual maintenance contracts for updates and technical support.
Expose scoring functionality via REST API that developers can integrate into their applications. Charge based on API calls with tiered pricing. Offer different pricing tiers for different scoring modes (basic heuristic vs. full panel+LLM).
💬 Integration Tip
Start with heuristic-only mode to establish baseline scoring, then gradually introduce panel and LLM judge features once users understand the basic workflow.
Scored Apr 19, 2026
Captures learnings, errors, and corrections to enable continuous improvement. Use when: (1) A command or operation fails unexpectedly, (2) User corrects Clau...
Self-reflection + Self-criticism + Self-learning + Self-organizing memory. Agent evaluates its own work, catches mistakes, and improves permanently. Use when...
A self-evolution engine for AI agents. Analyzes runtime history to identify improvements and applies protocol-constrained evolution. Communicates with EvoMap...
AI自我改进与记忆系统 - 解决'同类错误反复犯、用户纠正不长记性'的痛点。自动捕获错误、用户纠正、最佳实践，并转化为长期记忆。
Captures learnings, errors, and corrections to enable continuous improvement. Use when: (1) A command or operation fails unexpectedly, (2) User corrects Clau...
Self-improving agent system that analyzes conversation quality, identifies improvement opportunities, and continuously optimizes response strategies.