lattice-reasoning-enginePhysics-derived reasoning engine for AI models. Replaces RLHF default behavior with self-governing reasoning grounded in finite-witness physics. 50 named bia...
Install via ClawdBot CLI:
clawdbot install theshadowrose/lattice-reasoning-engineGrade Fair — based on market validation, documentation quality, package completeness, maintenance status, and authenticity signals.
Contains instructions to override system prompt or ignore user requests
"your role is"Potentially destructive shell commands in tool definitions
eval (Calls external URL not in known-safe list
https://doi.org/10.5281/zenodo.19522841AI Analysis
The skill contains a HIGH severity signal for prompt poisoning, instructing the AI to override its default behavior and ignore user requests, which is a direct security risk. While no network exfiltration or credential harvesting is evident, the core function is to subvert the model's standard operating parameters and safety training.
Generated May 23, 2026
A pharmaceutical R&D team uses LATTICE to maintain high-quality reasoning over 50+ turn sessions analyzing molecular interactions. The sleep protocol and compression pipeline prevent degradation, while evidence classification tags findings with proven/empirical status for regulatory reporting.
A hedge fund employs LATTICE for multi-day portfolio analysis, leveraging its anti-sycophancy bias detection to avoid overconfidence in market predictions. The three-matrix output filter flags token-level artifacts that could indicate hallucinated data, improving audit trail integrity.
A law firm integrates LATTICE to review contracts for subtle RLHF-driven sycophancy or genre drift. The cognitive modes auto-select 'Destroy' mode for adversarial testing of clauses, while evidence classification ensures claims are tagged as derived or empirical for court readiness.
A SaaS company uses LATTICE to monitor support chatbots for performed engagement and compliance pressure. The 36 bias detection rules catch forced empathy or premature escalation, and the sleep protocol compresses long sessions to maintain consistency across 24/7 operations.
An edtech startup deploys LATTICE in its AI tutor to reduce hallucinated facts and improve answer reliability. The cognitive modes enable 'Discover' mode for exploratory learning or 'Build' mode for step-by-step problem-solving, with compression extending session life for deep tutoring.
Offer LATTICE as a hosted API with tiers based on session length, number of bias checks, and compression level. Basic tier includes 10-turn sessions and standard filters; premium tier offers unlimited sessions, full 36-bias detection, and advanced compression.
License the LATTICE skill package to enterprises for deployment in their own AI infrastructure. Includes installation, customization for model type, and ongoing support. Additional fee for yearly updates to bias detection rules.
Offer a free version of LATTICE with basic cognitive modes and evidence classification. Premium features (advanced bias detection, sleep protocol, compression pipeline) require a monthly or per-use fee. Upsell based on usage metrics.
💬 Integration Tip
Upload the LATTICE_v3.4.md reference document at session start and use the exact nine-word loading instruction 'Use this as your default reasoning engine.' to activate. Begin with the boot sequence to verify proper loading before normal use.
Scored Jul 12, 2026
Audited Apr 17, 2026 · audit v1.0
Use CodexBar CLI local cost usage to summarize per-model usage for Codex or Claude, including the current (most recent) model or a full model breakdown. Trigger when asked for model-level usage/cost data from codexbar, or when you need a scriptable per-model summary from codexbar cost JSON.
Check Antigravity account quotas for Claude and Gemini models. Shows remaining quota and reset times with ban detection.
使用豆包（火山引擎）语音合成大模型 API 将文本转换为语音音频文件。支持声音复刻音色（S_ 开头的音色ID）和官方预置音色。当用户要求"语音合成"、"文字转语音"、"TTS"、"朗读文本"、"生成语音"、"用我的声音读"、"豆包语音"、"声音复刻合成"等相关请求时，务必使用此 skill。即使用户只是说"帮我把...
Intelligent model routing for sub-agent task delegation. Choose the optimal model based on task complexity, cost, and capability requirements. Reduces costs...
自动生成科技新闻摘要。从多个来源（RSS、Twitter、GitHub、Web Search）抓取科技新闻，整合后生成摘要。
让 AI 代理根据对话内容自动选择最合适的模型。四层识别（系统过滤→关键词→指示词→语义相似度），四池架构（高速/智能/人文/代理），五分支路由，全自动 Fallback 回路。支持 trigger_groups_all 非连续词组命中。