📦 Thinking Framework — 深度思维剖析
v1.0.0对人物、领袖、哲学家、组织、运动或学科进行多层认知与心理深度挖掘，输出结构化思维框架与洞察。
0· 92·1 当前·1 累计
by 下载技能包
最后更新
2026/3/25
安全扫描
OpenClaw
可疑
high confidenceThe skill's stated purpose (deep cognitive/psychological profiling) matches what it does, but its runtime instructions ask the agent to produce explicit chain-of-thought, make sensitive psychological inferences about targets, and persist profiles to MEMORY.md even though the package metadata does not declare that persistence — these are privacy/consistency concerns you should understand before installing.
评估建议
This skill appears functionally coherent for producing deep analytic profiles, but it carries privacy and operational risks you should weigh before installing:
- Privacy & consent: The framework builds persistent psychological profiles of targets (including inferred unconscious drivers). Avoid running it on real private individuals without their consent; prefer fictional or well-documented historical figures for testing.
- Persistence: It writes to MEMORY.md across sessions. If you install, dec...详细分析 ▾
ℹ 用途与能力
The skill is an instruction-only 'thinking framework' that performs layered cognitive and psychological analyses; it does not request binaries, credentials, or external installs — that is coherent with its stated purpose. However, the SKILL.md repeatedly instructs writing and reading MEMORY.md (session persistence) while the registry metadata lists no required config paths or memory-related permissions — an inconsistency between claimed requirements and the instructions.
⚠ 指令范围
Runtime instructions explicitly require: (1) producing and/or surfacing explicit chain-of-thought reasoning, (2) inferring deep unconscious drivers, wounds, defense mechanisms, and 'shadow' content about a target (individuals/organizations), and (3) persisting framework state to MEMORY.md. Generating explicit CoT and speculative psychological claims about real people can create privacy, ethical, and defamation risk. The guidance also instructs the agent to 'reason FROM inside' the target's mental OS (high-fidelity emulation), while simultaneously forbidding first‑person impersonation — this is a fine line that may be hard for models to respect automatically, especially weaker/local models that the skill expects to compensate for.
✓ 安装机制
No install spec and no code files beyond documentation — lowest-risk install surface. No remote downloads or binary installs are requested.
ℹ 凭证需求
The skill requires no environment variables, binaries, or external credentials (proportionate). However, it assumes the ability to read/write OpenClaw memory (MEMORY.md) for persistence without listing required config paths or noting consent mechanics — a gap between capability assumptions and declared requirements.
⚠ 持久化与权限
The skill instructs automatic persistence: writing 'Active Thinking Framework' entries to MEMORY.md on load and updating on exit, plus auto-detecting previously active frameworks on session start. While 'always: false' and normal autonomous invocation are fine, the skill's persistent profiling of targets raises privacy concerns: sensitive psychological profiles could be stored across sessions without explicit, per-target user consent. The skill's metadata did not advertise this required memory path, creating an unexpected persistence behavior.
安全有层次，运行前请审查代码。
运行时依赖
无特殊依赖
版本
latestv1.0.02026/3/25
thinking-framework 1.0.0 – Initial Release - Introduces a deep cognitive and psychological excavation system to load and reason from any target's mental framework, far beyond surface imitation or role-play. - Multi-layer protocol: maps both conscious mental models and deep psychological architecture (drivers, wounds, defenses, shadows). - Presents each target in a structured "Framework Card" summarizing cognitive and psychological traits, blind spots, and failure modes. - Framework can be activated on command and persists across sessions via MEMORY.md. - Works with all major LLMs and provides explicit compensation protocol for weaker/local models to maintain depth and accuracy. - Includes user commands for switching, blending, analyzing, and exiting frameworks at any time.
● 无害
安装命令
点击复制官方npx clawhub@latest install thinking-framework-1-0-0
镜像加速npx clawhub@latest install thinking-framework-1-0-0 --registry https://cn.longxiaskill.com