seo-geo-audit

Pass

Audited by Gen Agent Trust Hub on Nov 9, 2026

Risk Level: SAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [DATA_EXFILTRATION]: The skill is instructed to read local knowledge bases, specifically Obsidian vaults (searching for .obsidian directories) and markdown notes, to gather context on brand facts and keywords. This provides a mechanism for sensitive local data to be read into the agent's context.
  • [COMMAND_EXECUTION]: The skill utilizes subprocess.run in scripts/page_benchmark.py to execute internal scripts such as seo_audit.py and section_audit.py. It also instructs the agent to run shell commands for site crawls and content decay analysis using scripts located in sibling directories (e.g., ../seo-internal-linking/scripts/site_crawl.py).
  • [DYNAMIC_EXECUTION]: The script scripts/page_benchmark.py uses importlib.util to dynamically load the seo_audit module from a local file path at runtime.
  • [INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it fetches and processes untrusted data from external websites (HTML, headings, meta tags) which is subsequently processed by the LLM.
  • Ingestion points: scripts/seo_audit.py fetches raw HTML and metadata from URLs provided by the user or discovered during auditing.
  • Boundary markers: No explicit delimiters or instructions to ignore embedded commands are used when the model processes the collected site data.
  • Capability inventory: The skill can execute shell commands (subprocess.run), read local files (Obsidian vault), and write to local files (SEO action log).
  • Sanitization: The scripts parse the HTML into structured JSON, but they also extract raw text samples and headings which are passed to the LLM context.
  • [EXTERNAL_DOWNLOADS]: The scripts/seo_audit.py script performs network requests to scrape target websites. It explicitly disables SSL certificate verification by setting ssl.CERT_NONE, exposing the connection to potential Man-in-the-Middle (MitM) attacks.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 9, 2026, 11:06 PM