Thinking Protocol

Thinking-Claude inspired comprehensive thinking protocol with Verification Protocol and Confidence Scoring for OpenClaw agents

This skill is primarily a set of instructions telling the agent how to think and report confidence. It's coherent and low-risk from a technical-access perspective, but consider the following before installing: - Chain-of-thought exposure: the skill encourages streaming 'inner monolog' thinking blocks; if your deployment must avoid exposing internal reasoning (privacy, audit, or safety policies), do not enable or modify it to avoid publishing those blocks. - Verify provenance: registry metadata lists no homepage and the included package.json points to a GitHub repo; if you care about supply-chain provenance, check that repository and confirm the author before trusting the skill in sensitive environments. - Operational behavior: SKILL-howto mentions auto-injection into system prompts; confirm how your OpenClaw instance loads skills (it may or may not auto-enable this skill). Disable or restrict the skill if you want to avoid automatic behavioral changes across agents. - Metadata mismatches: minor version mismatch in package.json vs skill metadata and the howto wording vs registry flags — ask the publisher for clarification if you need tight change control. If you don't need chain-of-thought to be revealed to users, either avoid enabling the thinking-block output or adapt the protocol to produce summaries without verbatim internal monolog.

Type: OpenClaw Skill Name: thinking Version: 2.1.0 The 'thinking-protocol' skill bundle is a prompt-engineering framework designed to enhance AI reasoning through structured inner monologues, verification steps, and confidence scoring. The bundle consists entirely of Markdown instructions and metadata (SKILL.md, SKILL-examples.md) aimed at improving response quality and transparency; it contains no executable code, network requests, or instructions to exfiltrate sensitive data.