industry-scanner
Pass
Audited by Gen Agent Trust Hub on Jul 10, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes multiple internal Python scripts via shell commands to perform its scanning tasks. Evidence includes calls to
python3 skills/blog-scraper/scripts/scrape_blogs.py,python3 skills/reddit-scraper/scripts/search_reddit.py, and other subprocess calls documented across Phase 2 of the process. - [PROMPT_INJECTION]: A surface for indirect prompt injection exists because the skill ingests and processes large amounts of untrusted data from the public internet.
- Ingestion points: Untrusted content is gathered from Reddit, Twitter, Hacker News, blogs, and review sites as documented in
SKILL.mdPhase 2. - Boundary markers: The instructions do not specify the use of delimiters or 'ignore instructions' warnings when the agent processes this external data during Phase 3 and 4.
- Capability inventory: The agent has the capability to execute shell commands and write intelligence reports to the filesystem, providing a target for manipulation.
- Sanitization: No sanitization or filtering of the scraped content is mentioned before it is categorized and used to generate strategic opportunities.
Audit Metadata