programmatic-seo
Pass
Audited by Gen Agent Trust Hub on Nov 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting and processing data from external, potentially untrusted sources.\n
- Ingestion points: External data is ingested from Google autocomplete suggestions via suggestqueries.google.com, live SERP data via Firecrawl/WebSearch/WebFetch, and scraped entity-specific facts (pricing, features, quotes) during Phase 2.\n
- Boundary markers: The instructions do not define clear boundary markers, delimiters, or "ignore" instructions to prevent the agent from following malicious commands that might be embedded in the scraped text.\n
- Capability inventory: The skill possesses capabilities including shell command execution (python3, curl) and the ability to write files to the local directory (generating campaign files and site pages).\n
- Sanitization: There is no explicit requirement for technical sanitization or escaping of the external content before it is interpolated into the page templates or the answer-first block, though a manual quality sampling step is described.\n\n- [DYNAMIC_EXECUTION]: The skill incorporates dynamic execution patterns by generating and running local scripts during the build process.\n
- Evidence: In Phase 4, the skill describes creating and executing a scripts/generate_pages.py or .mjs script based on the dataset and template to render final HTML files. This script generation is derived from locally managed templates and data.\n\n- [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands for data collection and environment setup.\n
- Evidence: The skill uses curl to fetch data from suggestqueries.google.com and executes a local Python script (brand_context.py) to load project-specific context. These commands target well-known services and vendor-specific scripts.
Audit Metadata