upwork-apply
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill interacts with official APIs for several well-known services including Apify for web scraping, Anthropic for AI text generation, and Google Workspace (Docs, Sheets, Drive) for data storage and document creation.
- [INDIRECT_PROMPT_INJECTION]: The skill processes job listing data scraped from Upwork, which is untrusted external content, and uses it to generate AI-powered proposals.
- Ingestion points: Job titles, skills, and descriptions are read from a JSON file in
scripts/upwork_proposal_generator.pyafter being scraped byscripts/upwork_apify_scraper.py. - Boundary markers: Prompt templates in the proposal generator do not utilize specialized delimiters or instructions to prevent the LLM from following instructions potentially embedded within job descriptions.
- Capability inventory: The skill has permissions to perform network requests, manage Google Drive permissions (including public link sharing), and update spreadsheets and documents.
- Sanitization: Job-specific data is interpolated directly into system prompts for the Anthropic API without prior filtering or escaping of potential prompt injection sequences.
Audit Metadata