The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted enriched lead data from LeadGenius and synchronizes it directly into HubSpot CRM properties, creating a potential attack surface for indirect prompt injection.
Ingestion points: scripts/lg_full_to_hs.py fetches lead objects, including AI-generated justifications and cold email content, from the api.leadgenius.app endpoint.
Boundary markers: Absent. Data is mapped and written directly to HubSpot without delimiters or "ignore instructions" warnings.
Capability inventory: The skill possesses network writing capabilities (HubSpot Batch API) and local file system access.
Sanitization: Sanitization is limited to basic HTML tag replacement (e.g., <br> to \n) and does not filter for malicious instruction patterns within lead attributes.
[CREDENTIALS_UNSAFE]: The synchronization scripts contain a hardcoded CLIENT_ID ('6eceb43b-5c1a-4c4d-898c-36e67b24fdc6') identifying 'SCC Software
IT Automation'. Furthermore, scripts/lg_full_to_hs.py caches full lead details (PII) to a predictable local file path at /tmp/lg_full_leads_cache.json, which could lead to data exposure on multi-user systems.
[COMMAND_EXECUTION]: The skill instructions direct the agent or user to execute local Python scripts (scripts/lg_full_to_hs.py and scripts/lg_api_to_hs.py) to perform the data migration and schema setup.
[EXTERNAL_DOWNLOADS]: The skill documentation requires the installation of the requests and python-dotenv packages from the standard PyPI registry.

hubspot-leadgenius