product-description-seo
Pass
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted product catalog data which creates a surface for indirect prompt injection attacks.
- Ingestion points: Product names and existing descriptions are ingested from a JSON catalog file into the agent context in
scripts/analyze_catalog.py,scripts/extract_category.py,scripts/track_progress.py,scripts/update_catalog.py, andscripts/validate_json.py. - Boundary markers: The instructions in
SKILL.mdfor description generation (Phase 4) do not define clear delimiters or "ignore embedded instructions" warnings to separate untrusted catalog data from the agent's task instructions. - Capability inventory: The skill's scripts perform file system operations (reading/writing the JSON catalog and creating file backups using
shutil.copy2). TheSKILL.mdfile also provides instructions for the agent to execute shell commands using thegitandghCLIs. - Sanitization: The skill includes some sanitization logic;
scripts/check_quality.pyandscripts/validate_json.pyuse regular expressions to detect and block the presence of HTML tags or Markdown formatting in the generated descriptions.
Audit Metadata