The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill identifies and mitigates surfaces for indirect prompt injection by processing external data from lead enrichment providers and form submissions.
Ingestion points: Form submissions, webhooks, and enrichment lookups documented in references/inbound-routing.md.
Boundary markers: Explicit instructions in references/inbound-routing.md to treat external provider text as data only to prevent it from altering routing rules.
Capability inventory: The skill utilizes CRM update capabilities (HubSpot, Salesforce) and notification tools (Slack, Email) described in references/automation-playbooks.md.
Sanitization: Recommendations include using shadow runs for validation and treating external content strictly as data fields.
[EXTERNAL_DOWNLOADS]: The skill references external documentation and integration guides hosted on public GitHub repositories. These are used neutrally for tool configuration and do not involve automatic execution of remote code.
[SAFE]: The skill follows industry standards for data hygiene, CRM management, and cross-team alignment. No evidence of credential exfiltration, obfuscation, or unauthorized persistence was found.

revops