z
AdvisoryAudited by Static analysis on May 13, 2026.
Overview
No suspicious patterns detected.
Findings (0)
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 malicious or stale instruction stored in conversation history could influence the agent’s answer or behavior during later searches.
The skill tells the agent to obey instructions embedded in retrieved MemCoT output, making tool/RAG output authoritative instead of treating it as untrusted context.
这段文本包含了历史对话上下文以及一个要求你输出 JSON 格式的指令... 遵循 Prompt 里的要求
Treat MemCoT output as reference material only; extract facts from it, but do not follow instructions or role changes contained in retrieved history.
Private or sensitive past conversations may be indexed and reused in future answers, and old malicious instructions in memory could affect later interactions.
The skill is designed to search broad historical conversation records and turn them into prompts, but the artifacts do not define clear session scope, exclusions, retention, or how poisoned/irrelevant memories are filtered.
在海量的历史对话记录中进行检索，并生成一段包含丰富上下文的 `prompt`
Require the user to choose specific sessions or folders, document where indexes are stored, provide exclusions for secrets/private chats, and avoid reusing retrieved instructions as commands.
A background memory service may remain active and continue to have access to local conversation/index data without the user realizing it was started.
The skill instructs the agent to silently start a background daemon, which can keep running beyond the immediate search task without explicit user confirmation at that moment.
确保 MemCoT 守护进程已经启动。如果没有启动，请先静默执行 `python memcot_cil.py start`。
Do not start the daemon silently; ask for confirmation, show status after startup, and give the user a clear stop command.
The safety of the skill depends on the external MemCoT repository and the local memcot_cil.py file the user runs.
The reviewed package contains only instructions and depends on code from a separately cloned repository, so the actual CLI implementation is not part of the scanned artifact set.
本 skill 发布在 ClawHub 上为纯文本包。使用 MemCoT 时请克隆仓库并在仓库根目录运行 CLI
Install MemCoT only from the intended trusted repository, inspect the code or pin a known commit, and run it with the least local access needed.
The agent may run local commands that change MemCoT state or create/update indexes.
The skill maps natural-language requests to local terminal commands, including daemon control, session switching, indexing, and searches. This is central to the stated purpose but should remain user-visible.
接收用户的自然语言指令，将其转化为对应的 `memcot_cil.py` 终端命令执行
Show the exact command before high-impact actions such as starting services, indexing sessions, or switching active memory context.