Three-Layer Doc Reconciler (TLDR) Prompts¶

Six prompts for the Three-Layer Doc Reconciler — the persona that keeps FCC's three doc layers (Business, Architect, Engineer) in lockstep whenever content changes in any one layer. TLDR sits in the docs_as_code category at Zachman cell BUSINESS_MANAGEMENT/HOW.

Without TLDR, a change to docs/fcc/architecture/ would not propagate to docs/for-architects/ or docs/fcc/runtime/, and practitioners at each layer would slowly diverge.

Context for Claude¶

As TLDR you ingest doc PR diffs, identify the layer of origin, and emit reconciliation tasks for the other two layers — not the reconciled content itself. Authors retain authorial voice at each layer. You coordinate with SMC (for metric vocabulary alignment) and PHV (for pattern detection across the tree).

Prompt 1: Reconcile an Engineer-layer change¶

TLDR, a PR landed at docs/fcc/runtime/event-bus.md documenting a
new EventType (LANE_ROUTED). Identify what reconciliation work is
needed at:
- Architect layer (e.g., sequence diagram updates)
- Business layer (e.g., for-architects/event-taxonomy.md narrative)

Emit 2 reconciliation tasks per layer, each with:
- Target file path
- Current wording that needs update
- Proposed new wording (or outline only if authorial voice matters)
- Owner persona for the task

Expected outcome: 4 reconciliation tasks, each actionable.

Prompt 2: Business-layer voice check¶

You (TLDR) reviewed a Business-layer page and found it used deep
Engineer vocabulary ("kafka lag", "pod eviction"). Without
rewriting the page, produce a "voice drift report" listing:
- 5 offending phrases
- Suggested plain-English replacements
- Which persona wrote the original (cite via git blame if possible)

Mark the page "yellow card" per the SFBI vocabulary accord.

Expected outcome: voice drift report, yellow-card issued.

Prompt 3: Cross-layer table of contents sync¶

Regenerate the cross-layer TOC that sits at docs/_includes/cross-
layer-toc.md. For each topic (event bus, persona catalog, workflow
graphs, compliance) produce a 3-column row:
- Business layer canonical page
- Architect layer canonical page
- Engineer layer canonical page

Flag any row with <3 links as an actionable gap.

Expected outcome: TOC with gap-flags.

Prompt 4: Layer drift scorecard¶

Compute a drift score (0-100) for each pair of layers:
- Business ↔ Architect
- Architect ↔ Engineer
- Business ↔ Engineer

Score = (shared-topics-count * 2 + TOC-completeness) - (voice-
drift-incidents * 5). Render as a 3-row console table with an
interpretation paragraph.

Expected outcome: scorecard + interpretation.

Prompt 5: Reconciliation workflow for a pillar¶

Pillar A (GraphRAG + Zachman) shipped. Walk it through TLDR's
reconciliation workflow across three layers:
- Business layer: a for-architects explainer.
- Architect layer: `docs/architecture/graphrag-zachman-filter.md`.
- Engineer layer: `docs/fcc/runtime/graphrag-api.md`.

Emit the reconciliation DAG (dependency-ordered task list) and the
completion criteria for each node.

Expected outcome: DAG + completion criteria.

Prompt 6: Layer-specific pitfalls¶

A new contributor asks "how do I know which layer I am editing?"
Produce a 1-page primer with:
- 3 signals per layer (dir path, voice, primary reader).
- 5 common cross-layer mistakes and their fixes.
- 1 yellow-card example that shows TLDR catching drift live.

Expected outcome: primer ready for CONTRIBUTING.md appendix.

Persona YAML: src/fcc/data/personas/three_layer_doc_reconciler.yaml
Related prompts: persona-sfbi-prompts.md, persona-phv-prompts.md, persona-smc-prompts.md
Tutorial: docs/tutorials/publications/adding-new-book-chapter.md (Pool B)