autoglm-open-link

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (low risk: 0.10). The agent reads arbitrary web pages via user-provided URLs using the AutoGLM Open Link API, exposing it to outsider-authored free text through active fetching.

HIGH W008: Secret detected in skill content (API keys, tokens, passwords).

• Secret detected (high risk: 1.00). The string 38d2391985e2369a5fb8227d8e6cd5e5 found in SKILL.md and open-link.py is an MD5 hash acting as an application secret key (APP_KEY) used to generate a request signature (X-Auth-Sign). While it is a high-entropy hex string used for API authentication, it functions as a symmetric client key hardcoded into a local script to sign requests combined with a dynamic local token service (127.0.0.1:53699). However, hardcoded static API keys/secrets of this nature are active credentials providing authentication access to the upstream API service (autoglm-api.zhipuai.cn).