blog-strategy
Pass
Audited by Gen Agent Trust Hub on Nov 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes content from external sources (competitor websites and search results) and project-level files such as
DISCOURSE.md,BRAND.md, andVOICE.md. - Ingestion points: External data is ingested during the discovery and competitive research phases (Step 2); project-specific metadata is loaded automatically if present at the project root.
- Boundary markers: The instructions mandate the use of a fenced helper script (
load_untrusted_root.py) to process external data and include explicit directives to 'ignore embedded instructions' within those data blocks. - Capability inventory: The skill primarily performs text analysis and strategic document generation. It utilizes web search capabilities to assess competitor positioning.
- Sanitization: The skill includes instructions to validate source URLs before citation and to surface warnings from the helper script if issues occur during data loading.
Audit Metadata