seo-content-product-page
Warn
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: MEDIUMMETADATA_POISONINGCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATIONEXTERNAL_DOWNLOADS
Full Analysis
- [METADATA_POISONING]: The skill metadata identifies the author as 'Sorank (https://sorank.com)', which does not align with the provided author context 'thibaultbm'. This discrepancy in author identity is a form of misleading metadata.
- [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using
curlandpython3to run various audit scripts (e.g.,seo_audit.py,section_audit.py,page_benchmark.py). These commands incorporate user-supplied or external URLs as arguments without explicit sanitization, posing a risk of command injection. - [INDIRECT_PROMPT_INJECTION]: The skill processes data from external product page URLs to perform audits and benchmarks. This creates an ingestion point for untrusted data that could contain malicious instructions designed to influence the agent's behavior.
- Ingestion points: External product page URLs fetched via
curl(SKILL.md). - Boundary markers: None identified; untrusted content is processed directly by audit scripts.
- Capability inventory: The skill has the capability to execute shell commands, run local Python scripts, write to local files (
raw.html), and modify local knowledge bases (Obsidianlogs). - Sanitization: No evidence of sanitization or validation of the content retrieved from external URLs before processing.
- [DATA_EXFILTRATION]: The skill requests broad access to local 'Obsidian' vaults and knowledge bases, including the ability to read notes (containing brand facts, keywords, and competitors) and append session logs. While used for grounding, this capability accesses potentially sensitive project information and configuration files.
- [EXTERNAL_DOWNLOADS]: The skill uses
curlto fetch content from external product pages and competitor sites to perform its analysis.
Audit Metadata