skills/gpttang/skill-atlas/readgzh/Gen Agent Trust Hub

readgzh

Pass

Audited by Gen Agent Trust Hub on Jun 16, 2026

Risk Level: SAFEPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [PROMPT_INJECTION]: The skill ingests untrusted data from external WeChat articles, which constitutes a potential surface for indirect prompt injection attacks.
  • Ingestion points: External article content fetched from mp.weixin.qq.com URLs via the readgzh.read tool (SKILL.md).
  • Boundary markers: Absent; there are no instructions provided to the agent to treat the fetched article content as untrusted or to ignore embedded instructions.
  • Capability inventory: The skill utilizes network capabilities to communicate with the api.readgzh.site endpoint (SKILL.md).
  • Sanitization: No evidence of sanitization, filtering, or validation of the ingested article content was found in the documentation.
  • [DATA_EXFILTRATION]: The skill transmits user-provided WeChat URLs to an external API (api.readgzh.site). As this service belongs to the skill's vendor (GPTtang), this behavior is consistent with the primary purpose of the skill and the expected operational workflow.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 16, 2026, 11:09 AM
Security Audit — agent-trust-hub — readgzh