ai-architect

Design governed DISHA AI workflows with typed output validation and audit logging.

1|Updated Apr 8, 2026
One-click install
npx skills add https://github.com/Tashima-Tarsh/Disha --skill ai-architect
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: ai-architect
Source: https://github.com/Tashima-Tarsh/Disha/tree/main/.codex/skills/ai-architect
Command: npx skills add https://github.com/Tashima-Tarsh/Disha --skill ai-architect

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill addresses the risk of building DISHA AI workflows that lack governance, auditability, and safety guardrails, which can lead to unreliable outputs, unlogged critical decisions, and accidental exposure of sensitive data.

Core Features & Use Cases

  • Controlled Prompt Templates: Enforces use of standardized prompt templates to reduce AI output variability and ensure consistent behavior.
  • Deterministic Critical Decisions: Requires deterministic settings for high-stakes decisions to guarantee reproducible, verifiable outcomes.
  • Output Validation & Fallbacks: Mandates typed schema or guard function validation of model outputs, with fallback logic for invalid, unsafe, or unavailable responses.
  • Secure Audit Logging: Requires logging of decision metadata, explanations, fallback status, and request IDs while explicitly prohibiting logging of raw sensitive prompts or secrets.
  • Use Case: Use this Skill when building DISHA's governed AI components, such as lens analysis or policy gate evaluation, to ensure all outputs comply with DISHA's no-hallucination and evidence-led requirements.

Quick Start

Use the ai-architect skill to design a validated, fallback-enabled AI workflow for DISHA's cyber lens that logs all decision metadata without exposing sensitive input data.

Frequently Asked Questions about ai-architect

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I build governed AI workflows with reliable output validation?

To build governed AI workflows, apply standardized prompt templates, enforce typed schema validation for model outputs, and configure fallback logic to handle invalid or unsafe responses. This ensures reliable, verifiable, and safe AI behavior.

How do I prevent sensitive data exposure in AI audit logs?

Prevent sensitive data exposure in AI audit logs by configuring logging to capture only decision metadata, explanations, fallback status, and request IDs. Explicitly prohibit logging raw prompts, secrets, or sensitive input data.

What is the best way to ensure deterministic outcomes for critical AI decisions?

The best way to ensure deterministic outcomes for critical AI decisions is to require deterministic settings for high-stakes tasks. This guarantees reproducible and verifiable results, eliminating unpredictable AI hallucinations.

When do I need fallback logic in AI workflow design?

You need fallback logic in AI workflow design when model outputs fail typed schema or guard function validation. Fallbacks safely handle invalid, unsafe, or unavailable model responses, maintaining workflow reliability.

Can I use standardized prompt templates to reduce AI output variability?

Yes, you can use standardized prompt templates to reduce AI output variability. Enforcing controlled prompt templates ensures consistent behavior and satisfies evidence-led requirements for production-grade governed AI systems.