Start here — choose a role

If you are unsure, start with Option 1 (Practitioner).

Option 1 — Practitioner (daily AI use)
Reliable defaults for everyday AI work (drafting, summarizing, planning, analysis) without building systems.
Start with: Daily work guide
Add next (when needed): Prompt templates · Prompt components · CoVe procedure
Example: “Summarize these notes and list what is NOT proven under artifacts-only.”
Option 2 — Builder (agentic systems)
Building or changing systems where structure matters (architecture, boundaries, tool use, memory, orchestration).
Start with: Engineering quality gate
Add next (common stack): LLM memory boundaries · Prompt components · CoVe procedure
Example: “Review this repo snapshot for boundary violations and output a minimal changeset plan.”
Option 3 — Auditor (evidence-first verification)
Verifying claims, investigating failures, or enforcing citation discipline.
Start with: Fact-checking kit
Add next (when needed): Request web browsing · CoVe procedure · Confidence score (0–100)
Example: “Verify whether claim X is supported; fail closed if not.”
Option 4 — Writer/Publisher (publishable technical writing)
Turning AI-assisted drafts into publishable writing without unsupported claims.
Start with: Verify claims in writing
Add next (to stabilize semantics/reporting): Semantic accuracy gate · EGAM · Confidence score (0–100) · Request web browsing
Example: “Turn this draft into publishable content where every material claim is evidence-backed.”

Common mistakes