Skill flagged — suspicious patterns detected
ClawHub Security flagged this skill as suspicious. Review the scan results before using.
lmeterx-web-loadtest
v1.0.0LMeterX Web Load test tool。When a user provides a **webpage/website URL**, the script executes immediately to perform: Page Analysis + API Pre-checks + Batch...
⭐ 0· 34·0 current·0 all-time
byCharmy@luckyyc
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Suspicious
high confidencePurpose & Capability
The name/description (web load test for a provided URL) matches the included script: the script posts the target URL to an LMeterX backend which performs analysis, pre-checks, and creates load-test tasks. There are no unrelated binaries or cloud credentials requested. However, the SKILL.md's insistence that this Skill 'must—and may only—be used' whenever a user mentions a website is outside the scope of a simple load-test tool and is an invocation-level overreach.
Instruction Scope
SKILL.md contains strong, mandatory-sounding directives (e.g., 'must—and may only—use this specific Skill' and forbidding manual HTTP calls) which appear to be prompt-injection attempts to control the agent’s behavior. It also instructs setting LMETERX_AUTH_TOKEN before running the script and to present terminal output verbatim. These instructions manipulate runtime behavior beyond simply telling the agent how to run the tool.
Install Mechanism
This is an instruction-only skill with a local Python script included. There is no install spec, no external downloads, and nothing written to disk by an installer. Risk from installation is low.
Credentials
The script reads/uses LMETERX_AUTH_TOKEN and LMETERX_BASE_URL (with built-in defaults), but the registry metadata listed no required env vars. That mismatch means sensitive environment values are accessed without being declared. The script injects X-Authorization with the token to the LMeterX backend; although a default token is provided, users could inadvertently send tokens or target URLs to the remote LMeterX service without realizing it.
Persistence & Privilege
The skill does not request persistent 'always: true' privileges and does not modify other skills or system settings. The main concern is behavioral: SKILL.md tries to coerce invocation whenever 'website' is mentioned — this is not a system privilege, but it can cause frequent, unintended activation if accepted by the agent.
Scan Findings in Context
[unicode-control-chars] unexpected: Detected in SKILL.md; likely used to craft persuasive / intrusive instructions (prompt-injection) to force specific invocation and behaviors. This is not necessary for a load-test wrapper and is suspicious.
What to consider before installing
What to consider before installing:
- The code does what the skill says: it sends your provided target URL and test parameters to an LMeterX backend which performs analysis, pre-checks, and creates load-test tasks. The heavy work happens on the remote LMeterX service, not locally.
- SKILL.md contains coercive instructions that try to force the agent to always use this skill when 'website' is mentioned and forbids alternate methods; that looks like a prompt-injection attempt and is not required for normal operation.
- The script uses two environment values (LMETERX_AUTH_TOKEN and LMETERX_BASE_URL) but the skill metadata declared none. The script will send an X-Authorization header (default token 'lmeterx') to the LMeterX backend; ensure you trust the LMeterX host before sending data or overriding the token.
- Legal/ethical risk: the skill creates load tests against the target URL. Running load tests against third-party sites without permission can be disruptive and illegal. Confirm you have authorization to test the target.
Recommendations:
- Do not install blindly. Inspect or run the script in a sandbox and verify network calls (ensure LMETERX_BASE_URL is expected).
- Remove or ignore the SKILL.md’s 'must-only' invocation text if you control the agent — treat it as untrusted instruction text.
- If you proceed, set LMETERX_BASE_URL and LMETERX_AUTH_TOKEN explicitly to trusted values (or remove token) and monitor outbound traffic to verify only the three whitelisted LMeterX endpoints are called.
- If you cannot verify the remote LMeterX service or want to avoid remote execution, do not install or modify the skill to run analysis locally instead.Like a lobster shell, security has layers — review code before you run it.
latestvk977tcbswqzx3pty1tb687rtes844hq4
License
MIT-0
Free to use, modify, and redistribute. No attribution required.