job-posting-intent

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). The workflow searches LinkedIn job postings via Apify (harvestapi/linkedin-job-search), ingesting job descriptions and company details authored by random employers and posted publicly.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.70). The skill connects to https://rube.app/mcp and https://api.apify.com/v2 at runtime using hardcoded and environment-supplied credentials to execute remote jobs and build spreadsheets, but lacks official provenance establishing rube.app as a trusted first-party service.

HIGH W008: Secret detected in skill content (API keys, tokens, passwords).

• Secret detected (high risk: 1.00). The JWT tokens found in scripts/create_sheet_mcp.py, scripts/create_sheet.py, and scripts/search_jobs.py (e.g., eyJhbGciOiJIUzI1NiJ9...) are hardcoded credentials providing access to the Rube/Composio API (used as RUBE_TOKEN). These are real, active API/session tokens rather than placeholders or documentation examples, representing a leakage of sensitive credentials.