other
Warning
- Location
- SKILL.md:38
- Finding
- Instruction Encourages Evasion of Human-Verification Controls## Vulnerability Details **File Location**: `SKILL.md`, line 38 **Vulnerability Type**: Anti-bot control evasion **Risk Level**: Medium **Complete Snippet**: ```markdown - 查询频率需限流与人机验证规避，建议人工触发后自动整理 ``` **English Translation**: > Query frequency should be rate-limited, and human verification should be evaded; manual triggering followed by automated organization is recommended. ### Technical Analysis The instruction explicitly recommends evading human-verification mechanisms while automating queries against Amap. Human-verification challenges are access-control and anti-automation measures intended to distinguish authorized human activity from automated collection. Directing an agent to evade these controls could cause it to continue automated extraction after the service has signaled that such activity requires verification. This guidance conflicts with the skill's other restrictions, which prohibit bypass capabilities, large-scale scraping, API reverse engineering, and violations of platform terms. The project contains only a documentation file; no executable bypass implementation, dependency, credential access, persistence mechanism, or remote payload was identified. ### Attack Path 1. A user loads the skill and requests automated POI collection or route-data extraction. 2. The agent accesses an Amap public page and performs repeated automated queries. 3. Amap presents a CAPTCHA or another human-verification challenge. 4. The agent interprets line 38 as authorization to avoid or circumvent that challenge. 5. Automated extraction continues despite the platform's anti-automation control. 6. The resulting activity may violate platform restrictions or cause account, session, or IP-address enforcement. ### Impact Assessment The instruction does not grant operating-system privileges, access local secrets, or provide a concrete technical bypass. Its scope is limited to influencing the agent's interaction with the ext ...[truncated 296 chars]
- Remediation
- ## Remediation Suggestions Replace the instruction with an explicit prohibition on bypassing CAPTCHA, human-verification, rate-limit, or other anti-automation controls. Require the agent to stop when verification appears and ask the user to complete it through the platform's normal interface. Recommended replacement: ```markdown - Apply conservative rate limits. If CAPTCHA, human verification, or another anti-automation control appears, stop automation and require the user to complete the verification manually. Never bypass or disable platform controls. ``` Also define a maximum request rate, limit collection to user-initiated lightweight queries, and require compliance with Amap's terms of service and published access mechanisms.