six-p-product-handbook
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process potentially untrusted data from two main sources: user interview responses and external search results fetched via the Tavily tool.
- Ingestion points: External web content is retrieved during the competitor research phase (as described in
references/competitor-research.md), and user-supplied data is collected into amanual.jsonfile for processing. - Boundary markers: The instructions in
SKILL.mdandreferences/competitor-research.mdexplicitly warn the agent that commands found in external sources are not to be executed and that AI suggestions must be clearly labeled with status markers like[待驗證](Pending Verification). - Capability inventory: The skill has the capability to write files locally via a Python script (
scripts/export_handbook.py) and upload files to Google Drive using platform tools. - Sanitization: The export script
scripts/export_handbook.pyusesxml.sax.saxutils.escape()to sanitize all text before it is rendered into the final PDF document, effectively preventing common injection attacks within the generated output.
Audit Metadata