@davidme6

self-improving-enhancement adds chat logging, memory compaction, and pattern detection to a base self-improvement skill. claims of 65% speed gains and 73% fewer corrections lack validation. scripts exist but execution details are sparse.

self-learning-skill-v2 provides a framework for recording and analyzing errors with prevention measures, but lacks executable trigger phrases and concrete integration points. the error taxonomy and review cadence are well-structured, though the skill reads as aspirational rather than procedurally grounded.

smart model switcher v3 claims automatic multi-provider model selection across bailian, minimax, glm, and kimi with fallback logic. documentation is largely marketing narrative with no concrete implementation details, runnable procedures, or failure handling.

auth-guard enforces explicit user approval for external api operations across three modes (strict, whitelist, audit). covers configuration, audit logging, and integration patterns but lacks concrete trigger definitions and actionable failure recovery.

self-improving-v2 describes a tri-daily reflection agent with conversation archival and memory consolidation, but lacks concrete decision logic, error handling, and actionable procedural depth.

virtual-company simulates a 35-person organization with role-based sub-agents, memory persistence, and automatic rebuild on failure. documentation is extensive but procedures lack testable contracts and explicit failure recovery logic.

flash-company spawns temporary agent teams for one-off tasks using preset templates (dev, content, analysis, tech, design). adds persistent memory files in v1.1.0 but lacks actionable procedures, failure handling, and clarifies almost nothing about actual execution.

codex-delete-republish-test is a diagnostic skill for clawhub internal testing. it verifies whether a skill slug becomes immediately reusable after deletion and republication.

temp-test contains no actionable skill definition, intent specification, procedural steps, trigger phrases, or output contracts - it is repetitive filler text masquerading as documentation.

通用版Jarvis core，适配大多数平台和应用场景。主动智能助手核心技�?v3.1 - 从工具型助手到有灵魂伙伴的质变升级。在 v2.x 五大核心 + Phase 2 四项能力基础上，注入完整情感系统�? 种主情感光谱 / 三层架构（外�?内驱/内核�?/ 三模式情感倍率 / 6 层防崩溃防护 / 5 阶能力...

Universal Smart Model Switcher V3 - Multi-Provider, Multi-Model intelligent switching. Automatically selects the best model from ALL your purchased API plans...

Continuously identify knowledge gaps, proactively learn and apply new skills, regularly review progress, and iteratively improve technical and project abilit...

持续主动学习，深度分析问题根因，识别模式，横向扩展，迁移知识，避免重复错误，实现举一反三的自我迭代能力。

自我学习与迭代技能 v3.0.0 - 持续学习、错误预防、举一反三。包含能力评估、多渠道学习、进度追踪、错误学习系统、举一反三思维模型等核心功能。