company-current-gtm-analysis
Pass
Audited by Gen Agent Trust Hub on Mar 14, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it retrieves and synthesizes large amounts of untrusted content from the public web.
- Ingestion points: Phase 2 utilizes
WebFetchandWebSearchto collect data from company blogs, LinkedIn posts, review sites (G2/Capterra), and social media platforms (Reddit, Twitter). - Boundary markers: Absent. The instructions do not provide explicit delimiters or warnings to treat external content as untrusted data.
- Capability inventory: The skill possesses file read/write capabilities (reading
context.mdand writing the final report) and local command execution (running a Python script). - Sanitization: Absent. There is no logic provided to sanitize or filter potential malicious instructions embedded in the crawled content.
- [COMMAND_EXECUTION]: The skill executes a local Python script
scrape_linkedin_posts.pyviapython3to process LinkedIn profile data. - [EXTERNAL_DOWNLOADS]: The skill relies on external data sources for its core functionality, querying numerous domains including SimilarWeb, G2, Capterra, Greenhouse, and various social media platforms.
Audit Metadata