guardrails

白箱AGI架构探索：元认知（自我认知循环）、持续学习（知识飞轮）、世界模型（条件空间+语义时空图）、自我改进（自举纪律）、零LLM白箱管线与可审计信任护栏。

User-controlled policy and personalization runtime for DeepSeek Harness - enforce hard project constraints at the agent lifecycle boundary.

Commitment write-gate for AI coding agents: two-tier pre-execution policy (deterministic guard + LLM judge) with measured receipts. DeepSeek Harness plugin, engine-agnostic core.

The deny-by-default governance kernel for AI agents — permissions, L0–L4 guardrails, human approvals and an immutable audit chain as an Apache-2.0, framework-neutral library. Gate LangGraph / CrewAI / n8n / Dify or plain scripts in three lines.

DeepSeek Harness file-safety guardrail that limits AI coding-agent file edits per turn and prevents too many file changes, with call and UTF-8 byte budgets. 限制 DSH Agent 每轮修改文件数量。

Self-protection guardrails for the DeepSeek Harness: protected-path interception, backup-before-destroy, CLI self-recovery commands, and strict sandbox defaulting

DSH 外置状态机插件 · 幻觉抑制守卫 + 证据背书验证 + 上下文经济 · Structured state machine for long-horizon autonomous agents

Nine deterministic hooks that block a coding agent from weakening tests, swallowing errors, or stubbing type checks — each backed by a planted-failure test proving the guard can actually fire. Runs on Claude Code and DeepSeek Harness.

Indirect prompt-injection guard for DeepSeek Harness: taints tool output by origin, gates privileged calls that follow untrusted content, and refuses credentials heading off the machine.