Prompt Systems

The Six-Block Prompt System for Business Workflows

4 min read · Prompt Systems · Updated Jul 2026

The Six-Block Prompt System for Business Workflows
Six prompt blocks—Meta, Input, Output, Reasoning, Quality, Advanced—as a worksheet for repeatable business prompts.

The six-block prompt system—Meta, Input, Output, Reasoning, Quality, Advanced—for teams that need repeatable business prompts, not one-off chat wins.

Unstructured chat produces drafts and inconsistent customer-facing language. Sponsors do not fund one-shots; they fund outputs that survive review, audit, and handoff. The six-block prompt system is how Prompt Anatomy teaches teams to write that structure before they argue about models or tools.

If you need the operating stack—outcome through governance—start with Prompt Anatomy Foundations. If you need where reading vs practice live, see The Prompt Anatomy Ecosystem Map. This article defines the write layer.

The six blocks

Treat each block as a named contract, not a heading. Skip a block and you discover the gap in production review, not in the demo. Reviewers reject by missing block name—wrong voice, invented facts, or no refuse rules—instead of arguing about “tone.”

Block What you specify Failure if missing
Meta Role, situation, and purpose Generic tone; wrong stakeholder voice
Input Facts, sources, and constraints the model may use Invented numbers; silent policy violations
Output Format, sections, length, and delivery shape Unusable walls of text; wrong artifact type
Reasoning Steps or checks before the final answer Jump-to-conclusion drafts; weak trade-offs
Quality Pass/fail criteria and what to refuse Confident wrong answers ship unchecked
Advanced Tool or model settings that change behavior Drift when temperature, tools, or modes differ

The blocks are the anatomy of a designed prompt artifact. They are not a substitute for a prompt registry ID, an eval case set, or a RACI—those live in The Prompt Registry Playbook and governance posts.

Northline example — support reply

Northline B2B (composite) needed a support reply that named the product tier, refused inventing SLA dates, and ended with one clear next step. A six-block draft looked like this:

  • Meta: Senior support writer for Northline B2B; calm, precise; goal is a customer-ready email.
  • Input: Ticket summary, product tier, known outage window from the status page; do not invent dates.
  • Output: Subject line + three short paragraphs + one CTA bullet; no marketing upsell.
  • Reasoning: Confirm tier → check status facts → draft → self-check against Quality.
  • Quality: No invented SLAs; tier name exact; CTA is one action only; if facts missing, ask one clarifying question instead of guessing.
  • Advanced: Prefer deterministic drafting; no web browse unless status URL is provided in Input.

That structure turns “write a nicer reply” into an artifact reviewers can reject for a named reason. Paste the same six fields into the Six-Block Prompt Canvas for wiki and registry drafts. Practice loops that force fill-run-check on the same pattern live on promptanatomy.app after purchase—not as a slide deck.

Six-block vs RACE/TAG vs registry

Teams confuse three tools. Keep their jobs separate: writing structure, lighter task frames, and production ownership are not interchangeable.

Tool Job Use when
Six-block system Write a complete prompt with quality and settings Teaching structure; drills; capstone prompts
RACE / TAG Choose a lighter frame by task shape Communication drafts or role-action tasks—see Prompt Frameworks for Business
Prompt registry Version, own, and release production text Live workflows with eval-linked promotion

RACE and TAG can sit inside a six-block Meta/Output pair when the task is smaller. Registry rows should carry a stable ID and eval link even if the body was authored with six blocks. Do not declare a company-wide “winner” framework and skip owners.

Where practice fits

Shipping Prompt Anatomy describes the hub: plans, checkout, and module progress on promptanatomy.app. The Anatomizer on the marketing site teaches a five-part shape; Foundations teaches the six-layer ops stack; six blocks teach writing. Mixing those layers causes teams to buy training when they need a RACI—or to paste Anatomizer exports into production without owners.

Standardize team practice on promptanatomy.app when the same write-and-check loop is required. Keep governed workflows, eval hooks, and audit trails documented on this blog.

Guardrails

  • Completing modules or earning a practice certificate does not replace a workflow ID, eval gate pack, or change log on your systems.
  • Do not paste training copy into procurement decks; cite pass rate, cycle time, or incident cost on owned workflows.
  • Prefer one Northline-style worked example in your registry over twenty unlabeled chat wins.

Next: paste the blocks into the Six-Block Prompt Canvas. Shared drills: promptanatomy.app/#pricing.

On this page

Move from pilot to program

Structured training for teams implementing AI under real operational and compliance constraints.

Explore training

Continue learning

Step 4 of 7 in Prompt Systems · Full reading order

Go deeper

Template

Templates

AI Change Log Template

Copy-paste AI change log for prompt, context pack, model, and tool updates—with owner, eval evidence, and rollback registry pin.

4 min read · Templates · Apr 2026

FAQ

Is the six-block system the same as the Prompt Anatomy implementation stack?

No. The six blocks teach how to write a structured prompt. The implementation stack on Foundations is how teams operate AI—outcome, workflow, context, model step, evaluation, and governance.

Do teams need six-block training before using free blog playbooks?

No. Read frameworks on the blog first. Use training on promptanatomy.app when the team must practice the same write-and-check loop under feedback.

When should we use six-block vs RACE or TAG?

Use six-block when the prompt needs Quality and Advanced as named contracts. Use RACE or TAG for lighter task frames—they can sit inside Meta or Output; the registry still owns production IDs and eval links.