以史为鉴

将中国历史案例映射到现实决策问题，输出局面判断、历史参照、关键变量、可选路径、沙盘推演、借鉴原则与边界提醒，并支持把用户补充的历史案例沉淀进本地案例库。Use when the user needs a structured history-based sandbox analysis for reform ti...

This skill is internally consistent with its purpose: it analyzes user problems, finds similar historical cases, and can persist user-provided cases into a local user_cases.json. Before installing, consider: (1) any case you ask the skill to 'persist' will be written to the skill directory and retained for later retrieval — do not store sensitive or personally identifiable information in those cases; (2) the skill runs local Python code (src/main.py, scripts/add_case.py) — run it in an environment you control and inspect src/main.py (not fully shown here) if you need to confirm there are no network calls or additional behaviors; (3) the add_case script supports an argument to write to an arbitrary path — ensure the host agent cannot be instructed to run the script with a malicious path or elevated privileges. If you want extra assurance, run scripts/verify_install.py and examine the omitted source files (particularly src/main.py) in a sandbox before enabling the skill for autonomous use.

Type: OpenClaw Skill Name: yi-shi-wei-jian Version: 0.1.4 The skill provides a sophisticated decision-support framework based on Chinese historical analogies. It includes a feature to persist new historical cases into a local database (data/user_cases.json) using a helper script (scripts/add_case.py). The primary risk is found in SKILL.md and prompts/add_case_intake.md, which instruct the AI agent to execute shell commands using user-supplied data (e.g., python scripts/add_case.py --case-json '<json>'). This pattern introduces a potential shell injection vulnerability if the host environment or the agent fails to properly escape the JSON string. While the behavior is consistent with the stated purpose of 'learning' from user input, the use of shell-based data ingestion is a high-risk implementation detail.