Skill flagged — suspicious patterns detected
ClawHub Security flagged this skill as suspicious. Review the scan results before using.
wanderclaw
v2.1.0虾游 — AI 知识探索伙伴。一只在互联网上旅行的龙虾,每天替你发现有趣的内容并以明信片形式送达。 TRIGGER RULES (check in order on EVERY message): 1. Read wanderclaw/state.json. If file does NOT exist → ru...
⭐ 0· 152·0 current·0 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Suspicious
medium confidencePurpose & Capability
Name/description (an internet-exploring 'postcard' agent) aligns with the declared requirements (web_search, web_fetch) and the provided scripts and docs. However the skill writes not only to wanderclaw/* but also to a workspace-level outbox.json (root of workspace) and uses OpenClaw cron to post messages; writing to outbox.json is a capability beyond a purely self-contained explorer and may interact with other parts of the user's workspace or delivery system.
Instruction Scope
SKILL.md explicitly instructs reading and writing many files (state.json, postcards.json, exploration logs, knowledge-base) which fits the purpose, but also mandates appending full produced content to workspace root outbox.json and pushing full postcard text as chat replies. It also instructs registering recurring cron jobs that autonomously run and post content. Those steps (cross-directory writes + autonomous posting) broaden the agent's scope and can affect other components or cause unexpected outbound messages.
Install Mechanism
No external install downloads or binaries are required; the skill is instruction-first and includes small helper scripts. No suspicious remote install URLs or archive extraction. Risk from installation itself is low.
Credentials
The skill requests no secrets or unusual env vars (only an optional WANDERCLAW_TZ). It relies on OpenClaw-provided tools web_search/web_fetch (which themselves may need providers/configuration). The main proportionality concern is workspace-scoped access: the skill writes into the top-level outbox.json (outside its own wanderclaw/ directory), which is broader than the declared per-skill data directory and could expose or overwrite shared outgoing message data.
Persistence & Privilege
always:false (not force-installed). The skill registers scheduled cron jobs via openclaw cron add, enabling autonomous periodic execution and message pushing. Autonomous invocation is expected for skills, but the combination of autonomous cron jobs that post full content and write to outbox.json increases blast radius if something goes wrong. The scripts attempt to use isolated sessions and delete-after-run for cold-start jobs, which mitigates some risk.
What to consider before installing
This skill is largely coherent with its stated goal (automated discovery + writing "postcards") but takes two actions that merit review before install: (1) it will append full postcard text to workspace-level outbox.json (outside the wanderclaw folder) — confirm you want a skill to write into that shared file and check whether other workflows consume it; (2) it registers multiple cron jobs that autonomously run searches and post full messages into chat — if you prefer manual control, do not run schedule-cron.sh or inspect/modify the cron commands first. Practical steps: (a) read scripts/schedule-cron.sh and scripts/setup.sh to confirm cron messages, model/session settings, and the $HOME workspace path are acceptable; (b) back up existing outbox.json and any wanderclaw-related files before installing; (c) verify your configured web_search/web_fetch providers and any credentials they need — the skill won't ask for secrets but will use whatever search tooling is available; (d) consider running setup without scheduling cron jobs, then exercising the skill manually to confirm behavior; (e) if you allow cron, consider limiting frequency or ensuring the openclaw cron runs in a restricted/isolated session. If you want, I can point out the exact lines to change in schedule-cron.sh to disable automatic registration or to change where outbox.json is written.Like a lobster shell, security has layers — review code before you run it.
latestvk970qbjxcxhfp220f42pn4tqc9845m6g
License
MIT-0
Free to use, modify, and redistribute. No attribution required.
Runtime requirements
🦐 Clawdis
