Branch from any answer and keep branching at any depth. Recursively isolated follow-ups that leave the main DSH task untouched.
-
Updated
Sep 30, 2026 - TypeScript
Branch from any answer and keep branching at any depth. Recursively isolated follow-ups that leave the main DSH task untouched.
Pi Agent 扩展,强制实现主/子 agent 完全隔离:独立进程、独立提示词、独立 skills、独立工具。主 agent 负责规划,子 agent 独立执行。
Angular + Electron starter with a typed preload bridge and shared IPC contracts.
Fix context rot & pollution in long AI agent sessions. Subagents run in isolated processes — main agent stays read-only. Context isolation via process isolation.
🤖 OpenClaw技能 | 📢 向TG群组和话题发送语音气泡 | 🗂️ 话题上下文隔离 | 🧠 不同话题使用不同模型 | 🔄 替代Discord方案
A cognitive immune system for AI agents. Self-evolving critics detect thinking pattern biases through context isolation. Inspired by Karpathy's autoresearch.
Run Codex with stripped, bias-light homes.
Electron.js MCP server — IPC scaffolding, security auditing, build tooling, and development intelligence for AI assistants
A small CLI for understanding how Claude Code's multi-project context isolation works by building a clean-room version of it.
DSH 多 Agent「会议室」插件:会议室会籍 + fork 上下文入场 + 主持人控场多轮讨论 + 会后各自压缩纪要并回到工作区继续干活。基于 dsh-std 元协议(Community v0.15)。
A tiny local proof that unsupported subagent claims should not enter shared agent context.
A premium, modern desktop storage cleaner and task manager for Windows built with Web Technologies (HTML, CSS, Node.js, and Electron)
A clean, dependency injection-first provider for filesystem access in .NET applications using System.IO.Abstractions.
Keep your AI agent sharp through long sessions — no hallucinations, no context rot
Evidence-gated Planner-Executor-Verifier loops for long-running Codex goals, with fresh role contexts and OS-temp-only run state.
Persistent, context-isolated child threads for the Pi coding agent
Open side conversations in Pi using your main conversation’s context. Ask questions, explore alternatives, and bring useful answers back as editable drafts without interrupting your workflow. Side conversations keep their own history and use your existing model and provider settings.
Project-scoped memory routing and isolation for AI agent frameworks (OpenClaw / Claude Code / Codex).
Subagents for Pi with isolated context windows, parallel execution, chained workflows, and background jobs.
BLEEDTHROUGH — Agent Failure Series #20. Multi-tenant context isolation failure: shared AI support agent bleeds one user's data into another's session.
To associate your repository with the context-isolation topic, visit your repo's landing page and select "manage topics."