System Awakening
ReviewAudited by ClawScan on May 10, 2026.
Overview
This is a coherent learning assistant, but it asks the agent to automatically create persistent new skill files and memory from web-search results, which can affect future agent behavior without a clear review step.
Before installing, be comfortable with the skill creating persistent files under your local skills directory. Ask the agent to show you the full generated plugin before saving it, and avoid using execution mode for real system, account, or file changes unless you explicitly review and approve each action.
Findings (3)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
A generated skill could remain active across future sessions and influence later agent responses or actions, even though the user may not have separately reviewed the final file.
The skill instructs the agent to automatically write a new persistent skill/plugin file into the local skills directory after resource filling, which can change future agent behavior.
触发时机：第二轮资源填充完成后，自动执行。... 文件路径：`~/.workbuddy/skills/[topic-slug]-talent/SKILL.md`
Require explicit user approval before writing any generated skill file, show the full generated content first, and provide a clear uninstall or rollback path.
Bad or manipulated search results could be saved into a generated skill and reused later as if they were trusted guidance.
The workflow derives skill content from web search results and then stores it as a reusable skill file, creating a path for untrusted retrieved content to become persistent future context.
从搜索结果中提取技能节点... 将完整天赋技能树写入独立的 Skill 文件
Treat search results only as references, avoid saving untrusted instructions into generated skills, and require the user to inspect generated content before it is reused.
If used with powerful tools, the agent might move from advice into performing file, code, server, or account changes under a broad 'execute' instruction.
The execution mode is purpose-aligned, but it broadly tells the agent to perform tasks based on generated skill capabilities, without spelling out extra approvals for high-impact actions.
如果Y在技能能力范围内 → 确认需求 → 执行 → 返回结果
Use execution mode for drafts and guidance by default, and require separate confirmation before running commands, editing important files, changing servers, or using accounts.