The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from LinkedIn posts to generate intent signals, creating a potential surface for indirect prompt injection.
Ingestion points: LinkedIn post text is ingested in scripts/signal_scanner.py via Apify scraping.
Boundary markers: The LLM prompt specifically constrains the model to "Return ONLY a single integer from 1-10", which limits the impact of adversarial content.
Capability inventory: The skill has the ability to insert records into the signals table and update lead_status for personas in Supabase.
Sanitization: The script requires a --dry-run and explicit user approval before performing any database writes, ensuring human oversight of detected signals.
[EXTERNAL_DOWNLOADS]: The skill utilizes external data collection via Apify actors to perform job searches and profile scraping.
Evidence: scripts/signal_scanner.py invokes harvestapi/linkedin-job-search, harvestapi/linkedin-profile-posts, and supreme_coder/linkedin-profile-scraper. These are standard third-party tools provided through the Apify platform, a well-known scraping service.

signal-scanner