geo-visibility

Warn

Audited by Gen Agent Trust Hub on Oct 29, 2026

Risk Level: MEDIUMDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONMETADATA_POISONINGDATA_EXFILTRATION
Full Analysis
  • [DYNAMIC_EXECUTION]: The skill references a Python script at scripts/seo_audit.py that is not part of the analyzed files. Invoking external or missing scripts can lead to the execution of unverified code.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes content from external websites and local markdown notes, making it vulnerable to indirect prompt injection.\n
  • Ingestion points: Local Obsidian notes and external HTML content fetched via curl.\n
  • Boundary markers: The instructions do not specify delimiters or safety warnings for the agent when handling this data.\n
  • Capability inventory: The agent can execute scripts, perform network requests, and write to local files.\n
  • Sanitization: No sanitization of the fetched data is mentioned.\n- [DATA_EXFILTRATION]: The skill requires reading sensitive local data, including the contents of Obsidian vaults. Combined with the agent's ability to use curl, this provides a potential path for data exposure or exfiltration.\n- [METADATA_POISONING]: There is a discrepancy between the author listed in the skill's frontmatter ("Sorank") and the platform-provided author ("thibaultbm"), which is considered deceptive metadata.\n- [COMMAND_EXECUTION]: The skill instructs the use of curl for fetching raw HTML, which is a network command that could be misused if directed toward sensitive endpoints.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 29, 2026, 12:06 AM