Resources — Kindo Training Portal

📖 Official Documentation

🎯 Use Case Templates

Cybersecurity Workflows

Firewall Rule Optimizer Categorize firewall rules for cleanup: risky, over-permissive, unused, redundant · Exercise A

📋

NIST CSF Compliance Mapper (Pathfinder) Map evidence documents to NIST CSF controls and evaluate compliance status · Exercise B

🚨

Alert Triage Analyze security alerts from CrowdStrike, Splunk, or other tools — prioritize by severity and context

🔍

Vulnerability Reporting Aggregate vulnerability scan results and generate prioritized remediation recommendations

🔎

Threat Hunting Query integrated tools for indicators of compromise with structured reporting output

Additional Use Cases

Incident response automation

IAM access reviews

Cloud infrastructure audits

Compliance reporting (ISO 27001, SOC 2, PCI DSS)

Log analysis & anomaly detection

🔌 Integration Guides

Integrations relevant to Deloitte's stack (must be pre-provisioned before the pilot session):

🦅

CrowdStrike Falcon Threat intelligence, endpoint detection, and alert enrichment

🎯

Jira Auto-create remediation tickets from agent findings

🛎

ServiceNow Incident management and workflow automation integration

🔦

Splunk Log analysis, SIEM integration, and alert aggregation

🔗

All Supported Integrations Full integration catalog on docs.kindo.ai

⭐ Best Practices Quick Reference

Prompt Engineering

Role specification

State the agent's role explicitly: "You are a senior security analyst reviewing…"

Output format

Define expected output format explicitly — tables, JSON, structured reports. The model follows them well.

Constraints

Specify boundaries: "Only report Critical and High severity findings" or "Choose the highest-risk category."

Confidence levels

Add HIGH / MEDIUM / LOW confidence to outputs — makes findings actionable (auto-fix vs. manual review).

Knowledge Store Tips

Attach only files relevant to the agent's task
Prefer structured formats (Markdown, CSV, JSON) over unstructured text or screenshots
Keep reference files current with a regular review cadence
For large documents, consider breaking them into focused chunks by topic

Agent Construction

Give each step one job. If a step analyzes, prioritizes, writes a report, and files a ticket — it's doing too much.
Name steps descriptively. "Fetch Open Vulnerabilities" is more maintainable than "Step 1".
Use the right step type. LLM Steps for reasoning, API Action Steps for deterministic reads/writes, Action Steps for integrated product actions.
Test iteratively. Build one step, test it, then add the next. Don't try to build the entire agent at once.
Start simple. A 2-step agent that works is better than a 10-step agent that doesn't.

💬 Support & Contact

Kindo Account Team

Ron Williams · Tony Wong · Bryan Vann · Sean Walker · Joana Dias

Deloitte Coordinators

Smriti Kewlani · Manoj (pilot scheduling)