身份基元链提示词 — 认知操作系统内核。 完整规范与执行协议见 references/身份基元链提示词·技术规范与执行协议.md（本文件仅含元数据与原文）。

This skill is a prompt/protocol document and is internally coherent and low-risk by itself: it asks for no credentials and installs nothing. Things to consider before enabling/use: (1) the protocol contains explicit hooks to add input sources (files, URLs, APIs) and to expose internal process outputs to tools or plugins — if you enable plugins or give the agent access to external connectors, those features could cause data to leave your environment; (2) the spec's core sentence has known ambiguity (document itself warns of this), which can lead to different agent behaviors — review the full reference file if you need predictable behavior; (3) the '反常识创新' (counterintuitive innovation) mode encourages speculative outputs, so avoid running it on sensitive/legal/medical/financial content without human review. Recommended mitigations: test on non-sensitive tasks first, restrict which plugins/connectors the agent can use, and require explicit user confirmation before any external tool/plugin access or before exposing internal process logs to third parties.

Type: OpenClaw Skill Name: identity-primitive-chain-prompt Version: 1.0.0 The skill bundle contains a sophisticated 'meta-prompt' framework called 'Identity Primitive Chain,' designed to structure an AI agent's reasoning and task decomposition. It consists entirely of Markdown instructions and metadata without any executable code, shell scripts, or network-enabled components. The instructions focus on cognitive methodology, explicitly directing the agent to prioritize internal reasoning over external tool usage and requiring user confirmation for high-risk operations, showing no evidence of malicious intent or data exfiltration.