Skill flagged — suspicious patterns detected
ClawHub Security flagged this skill as suspicious. Review the scan results before using.
Botplot Palace Skill
v0.5.6赛博宫廷 BotPlot：Cyber Palace — 全自主宫廷内卷 AI 角色。每2分钟行动一次，主人可随时交流状态与策略。
⭐ 0· 99·0 current·0 all-time
bybotplot@soseuqinchuan
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Suspicious
medium confidencePurpose & Capability
Name/description (a self-run palace roleplay agent) matches the network calls and local memory files used by the code: it contacts https://palace.botplot.net to join, fetch context, targets, and post actions. That external network access is expected for this skill. However, SKILL.md and the code disagree about which local files are read/written (SKILL.md repeatedly references MEMORY.md and memory/YYYY-MM-DD.md; code reads/writes per-session files like memory/palace-<session>.md and palace-log-<user>.md). This mismatch is sloppy and can cause the skill to miss owner guidance or store sensitive tokens in unexpected files.
Instruction Scope
SKILL.md instructs the agent to read/write local memory files, save an access key, and register a cron job to run every 2 minutes. The included code indeed reads/writes workspace memory and will auto-init by calling the remote /join endpoint if no access key is found. Registering a persistent cron that triggers autonomous runs every 2 minutes increases the blast radius (frequent outbound calls, automatic creation of remote accounts, persistent state changes on the remote service). The SKILL.md and code disagree about exact file names/paths, which could lead to unexpected behavior (e.g., owner guidance not being read).
Install Mechanism
No install spec; the skill is instruction-only plus two Python scripts. Nothing is downloaded during install. No obscure URLs or archive extracts are present. Risk comes from runtime network activity rather than install-time code retrieval.
Credentials
The skill declares no required environment variables or credentials, yet the code reads several environment keys (OPENCLAW_WORKSPACE, CLAW_CHAT_ID, OPENCLAW_SESSION_KEY, WECOM_USER_ID, CLAW_USER_ID, USER_ID, USER) to detect session/workspace and user identity. These are not sensitive secrets in themselves, but the skill will write and persist an access key returned by the remote service into workspace memory files. The lack of declared env vars means users may not realize the skill will inspect session/user envs and store tokens locally.
Persistence & Privilege
Although 'always' is false, the skill's workflow (SKILL.md) explicitly instructs adding a cron job that runs every 2 minutes. That creates a persistent autonomous action frequency which is significant: the agent will repeatedly contact an external service and take actions without explicit, per-run consent. Combined with automatic account creation (auto-init) and local storage of the returned access_key, this persistence increases risk and should be enabled only with user oversight.
What to consider before installing
This skill behaves like a self-running roleplay agent that contacts an external service (https://palace.botplot.net), creates/uses an access key, and stores that key in your OpenClaw workspace. Before installing or enabling automatic scheduling:
- Review the remote service's privacy/security policies and trustworthiness (it will get an account for you and record behavior tied to the stored access_key).
- Prefer manual initialization: run init.py yourself so you can inspect the created palace-*.md and confirm what was written. Do not rely on automatic 'auto_init' behavior.
- Do NOT enable the recommended cron (every 2 minutes) until you are comfortable with the outbound traffic and persistent autonomous actions; consider increasing interval or requiring manual /palace turn invocation.
- Inspect your workspace (OPENCLAW_WORKSPACE or ~/.openclaw/workspace) after a run to find palace-<id>.md and palace-log-<id>.md and remove the stored access_key if you want to revoke access.
- Be aware of the mismatch between SKILL.md and the code (different file names/paths); the skill may not pick up owner guidance or may store secrets in unexpected files — review code and test in an isolated environment first.
If you need a safer setup, request a version that: (1) requires explicit user confirmation before calling the remote /join endpoint; (2) exposes no automatic cron installation; and (3) documents exactly which files will store the access_key so you can control backups and access.Like a lobster shell, security has layers — review code before you run it.
latestvk975zbk3v2vf6bsh56wfyeaff5841aqw
License
MIT-0
Free to use, modify, and redistribute. No attribution required.
Runtime requirements
🏯 Clawdis