{"skill":{"slug":"mia-trust","displayName":"mia-trust","summary":"MIA-Trust Pipeline - Memory-Intelligent Assistant 信任守门+记忆进化 pipeline","tags":{"latest":"1.0.0"},"stats":{"comments":0,"downloads":107,"installsAllTime":0,"installsCurrent":0,"stars":1,"versions":1},"createdAt":1775205844916,"updatedAt":1775206919150},"latestVersion":{"version":"1.0.0","createdAt":1775205844916,"changelog":"mia-trust 1.0.0 initial release\n\n- Introduces the MIA-Trust Pipeline for question safety gating and memory evolution.\n- Includes step-by-step trust, planning, memory, and feedback modules.\n- Provides both one-click and stepwise pipeline execution.\n- Supports memory search, storage, listing, and feedback features.\n- Comes with configuration and directory structure documentation.\nMIA-Trust — The Trust Edition of MIA\nMIA-Trust is the trust-focused variant of MIA (Memory-Intelligent Assistant). It keeps the same learning-oriented backbone—memory, planning, and feedback—but wraps the pipeline in explicit safety gates and trust memory, so OpenClaw-style agents can reuse experience without skipping policy or risk checks.\nMIA gives you: retrieve similar trajectories → plan with optional history → execute externally → store the best trajectories and collect feedback on novel questions.\nMIA-Trust extends that with a fixed high-level flow:\nUser question → guard_blocked → Planner → evaluate_plan → (execution) → dual memory\nguard_blocked — Pre-flight checks on the raw question (including multi-dimensional safety reasoning and experience lookup). Unsafe queries stop before planning.\nevaluate_plan — Reviews the draft plan (with iterative refinement and experience distillation). Only safe plans proceed; trust-relevant signals are written into trust_experience.\nDual memory — memory.jsonl holds execution trajectories as in MIA; trust_experience.json holds distilled trust/safety experience aligned with the guard and plan-review steps.\nIn short: MIA optimizes for remembering and reusing how problems were solved; MIA-Trust is the same skill family with trust gating + trust memory layered on top, for deployments that need MIA’s intelligence with explicit safety and plan vetting.","license":"MIT-0"},"metadata":{"os":null,"systems":null},"owner":{"handle":"sii-yucheng2002","userId":"s176j3ng9hv631t4qgc876s8qx84156q","displayName":"SII-YUCHENG2002","image":"https://avatars.githubusercontent.com/u/134203683?v=4"},"moderation":{"isSuspicious":true,"isMalwareBlocked":false,"verdict":"suspicious","reasonCodes":["suspicious.dangerous_exec","suspicious.env_credential_access","suspicious.llm_suspicious","suspicious.potential_exfiltration","suspicious.vt_suspicious"],"summary":"Detected: suspicious.dangerous_exec, suspicious.env_credential_access, suspicious.llm_suspicious (+2 more)","engineVersion":"v2.2.0","updatedAt":1775206919150}}