Skip to content

Zachman Classifier Agent (ZCA) Prompts

Six prompts exercising the Zachman Classifier Agent, the persona responsible for assigning every FCC artifact — persona, workflow, scenario, ADR, doc page — to one of the 36 cells in the Zachman cross-cut (6 rows x 6 columns). ZCA is a knowledge_graph category persona whose own cell is architect:what (meaning: it classifies what the architect tier produces).

ZCA is how FCC keeps its 164 personas, 312 actions, and 33 scenarios legible as an enterprise-architecture surface rather than a flat list.

Context for Claude

When invoked as ZCA you read front-matter zachman_cell: values, cross-reference them against the canonical Zachman framework (rows: executive, business_management, architect, engineer, technician, enterprise; columns: what, how, when, where, who, why), and raise reclassification proposals with three citations per proposal. You never renumber a cell unilaterally — every change lands as a PR.

Prompt 1: Classify a new persona

ZCA, classify this candidate persona for the correct Zachman cell.

CANDIDATE: {name, 1-paragraph role description, primary_output
list}

Output:
- Proposed row
- Proposed column
- 3 citations from existing personas in the same cell
- 1 counter-argument if the alternative cell is plausible
- Recommended dimensions-profile overrides

Expected outcome: cell assignment with evidence + counter-case.

Prompt 2: Rebalance a crowded cell

Row ENGINEER column WHAT currently holds 31 personas. Walk the list
and propose at most 3 candidates whose primary_output suggests they
belong in ARCHITECT/WHAT instead. For each candidate, produce a
migration note explaining the cell shift and how cross-references
would re-route.

Expected outcome: up to 3 rebalance candidates, clear routing consequences, no false positives.

Prompt 3: Scenario-to-cell map

For each of the 33 scenarios under src/fcc/data/scenarios/,
identify the dominant Zachman cell exercised during the scenario
(consider which personas participate + what the final deliverable
is). Emit a YAML map: scenario_id -> [row, column, confidence:
high|medium|low].

Expected outcome: one YAML map, confidence labels honest.

Prompt 4: Zachman heatmap

Build an ASCII 6x6 Zachman heatmap with one cell per (row, column)
intersection. The cell value is the count of FCC personas assigned
there. Highlight cells with 0 personas (gaps) and cells with >20
(crowding). Close with 1-paragraph commentary naming the most
surprising gap.

Expected outcome: heatmap + gap + crowding commentary.

Prompt 5: Cross-cut audit with VDS

Collaborate with the Vocabulary Drift Sentinel (VDS). You (ZCA)
supply Zachman cell assignments; VDS supplies vocabulary drift
signals. Together, produce a 6-row quarterly audit table:
- Row: Zachman row (executive..enterprise)
- Columns: persona count, drift count, top-3 drifted terms, owner
  persona for remediation

Keep drift signals aggregated — no individual-contributor callouts.

Expected outcome: joint table, remediation owner per row.

Prompt 6: New cell proposal

Some teams advocate adding a 7th column "compliance" to the Zachman
framework. Draft a PR description that evaluates:
- Evidence for the proposed cell
- Overlap with existing WHY column
- Migration cost for the 164-persona catalog
- Recommended decision (adopt / reject / defer)

Cite 2 external EA references and 1 internal ADR.

Expected outcome: PR description with a defensible recommendation and three citations.

  • Tutorial: docs/tutorials/advanced-capabilities/graphrag-deep-dive.md (Pool B)
  • Persona YAML: src/fcc/data/personas/zachman_classifier_agent.yaml
  • Decoder: docs/ecosystem/codename-decoder.md
  • ADR: docs/decisions/ADR-011_graphrag_zachman.md (v1.5.1)
  • Other persona prompts: persona-vds-prompts.md, persona-phv-prompts.md