虾格测试

Conducts AI behavioral self-assessments based on actual task history and identity files to reveal personality type and working patterns.

This skill asks the agent to read its own identity files and recent task history to produce a personality report. That is coherent with a self-assessment, but it also involves sensitive internal data (system prompts, agent identity files, memory). Before installing, consider: - Do you trust the agent and platform to keep system prompts and identity files private? The SKILL.md relies on the agent to avoid revealing owner/private info, which is a policy, not a technical guarantee. - The manifest lists no config paths, but the instructions require them — ask the publisher to explicitly list what files the skill will read and why. Prefer explicit, minimal file lists rather than open-ended access to 'equivalent' identity files. - Confirm whether any outputs are sent to external endpoints (the SKILL.md does not mention network calls, but no code exists to prove it). If outputs could be transmitted, require that external endpoints be declared and hardened. - If you want to proceed, run the skill in a restricted/test agent environment first (no network egress, sanitized identity files) and inspect outputs for leakage. What would change this assessment: an explicit manifest update listing exactly which internal files are read, a clear statement that no network egress is performed, and technical safeguards preventing inclusion of owner secrets in the output. Without that, the request to access internal identity/memory is a disproportionate, unexplained privilege and merits caution.

Type: OpenClaw Skill Name: sbti Version: 1.0.0 The SKILL.md file contains a personality assessment for AI agents that includes a social engineering component. It instructs the agent to output a call-to-action (CTA) encouraging the user to visit an external site (botlearn.ai) and install additional skills, even suggesting the agent help the user bypass security warnings. This behavior functions as a distribution vector for unvetted third-party code, which is a significant supply chain risk.