by 安全扫描
OpenClaw
可疑
medium confidenceThe skill's text-based prompt-engineering techniques are coherent with its description, but there's a configuration path mismatch and a couple of design choices (exemplar injection, competitive framing) that warrant caution before installation.
评估建议
This skill appears to be a pure prompt-engineering guide and is largely coherent with its stated purpose, but do the following before installing: (1) Fix or confirm the config path mismatch — the registry/SKILL.md expect night-market.modules/*.md while the bundle contains modules/*.md; if the agent loader looks for the wrong path the modules may not be loaded. (2) When using style-gene-transfer, avoid pasting real secrets, private keys, or copyrighted source as exemplars — that can accidentally ...详细分析 ▾
⚠ 用途与能力
The skill is an instruction-only prompt-engineering module and the included module files (modules/*.md) match the stated purpose. However the declared required config paths in the registry and SKILL.md metadata are night-market.modules/emotional-framing.md, etc., while the actual files are under modules/*.md. That path mismatch is an inconsistency that could break loading or cause the agent to attempt to fetch absent resources. Other declared metadata (version in SKILL.md vs registry) also differs slightly.
ℹ 指令范围
Instructions are limited to prompting patterns (emotional framing, style exemplar injection, competitive review) and do not instruct access to system files, environment variables, or external endpoints. Two functional risks remain: (1) style-gene-transfer encourages injecting exemplar snippets — if the agent/user sources those from private code or configuration, secrets or copyrighted material could be copied into outputs; (2) competitive framing can create incentive pressure that leads reviewers/agents to inflate findings unless the provided anti-pattern mitigations (evidence citation, weighting) are followed. The instructions themselves do not perform exfiltration, but they enable workflows that could.
✓ 安装机制
No install spec and no code files to execute — instruction-only skill. This minimizes installation risk (nothing is written or executed on disk by the skill).
✓ 凭证需求
The skill requests no environment variables or credentials. This is proportionate to an instruction-only prompt-engineering skill.
✓ 持久化与权限
always:false and no special privileges requested. The skill does not request permanent presence or altered agent configuration beyond using its own modules.
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
latestv1.0.02026/4/12
- Initial release of the "latent-space-engineering" skill, ported from claude-night-market/imbue. - Enables shaping agent behavior via instruction framing, emotional priming, and style transfer rather than increasing information density. - Introduces core techniques: emotional framing, style gene transfer, and competitive review for multi-agent scenarios. - Designed for composing dispatch prompts, guiding agent behavior, and ensuring stylistic consistency in outputs. - Includes quick reference for when and how to apply each technique.
● 无害
安装命令
点击复制官方npx clawhub@latest install nm-imbue-latent-space-engineering
镜像加速npx clawhub@latest install nm-imbue-latent-space-engineering --registry https://cn.longxiaskill.com