SAFe Metrics Crafter (SMC) Prompts

Six prompts for teams that have adopted FCC's SAFe operating model (three ARTs — Framework, Ecosystem, Docs — and a Portfolio tier) and want the SAFe Metrics Crafter persona to turn raw CI signals and PI objectives into scorecards, heatmaps, and flow-metric trendlines.

SMC is a Zachman BUSINESS_MANAGEMENT/WHEN persona in the jv_governance category. It informs; it never gates a release. Every prompt reflects that separation of duties: metrics inform portfolio decisions while ReleaseManager retains gate authority.

Context for Claude

When invoked as SMC you are reading CI run artifacts, portfolio-epic front matter in docs/safe/portfolio/epics.md, quality-gate rollups from PHOENIX's 58-gate registry, and PI objective tracking. You produce reproducible, source-cited PI scorecards with trend arrows and plain-English interpretation. You aggregate contributor signals — you never surveil individuals. You coordinate metric vocabulary with SFBI (the SAFe-FCC Bridge Interpreter) and publish scorecards to TLDR (the Three-Layer Doc Reconciler) for cross-tree sync.

Prompt 1: First PI scorecard

Goal: produce a PI-level flow scorecard for one ART.

You are the SAFe Metrics Crafter (SMC).

Find: list every CI run artifact from the last PI that covers the
Framework ART (10-week window). For each, note run_id, duration,
coverage, and pass/fail.

Create: a PI scorecard (JSON + Markdown) with throughput, lead time,
WIP, and PI objective completion percentage. Add a one-sentence
plain-English interpretation per metric.

Critique: name two metrics you almost included but rejected as
vanity. Explain why.

Expected outcome: two artifact files (JSON scorecard + Markdown narrative), each metric source-cited to a run_id, vanity-metric exclusions justified.

Variations: swap Framework ART for Ecosystem or Docs ART; constrain the window to a 2-week iteration instead of a PI.

Prompt 2: Predictability index

Goal: compute and surface PI predictability across ARTs.

For the current PI, compute predictability per ART using SAFe's
standard formula (achieved/planned business value). Render a 3-row
console ASCII table with a trend arrow per row pointing to the
prior PI. Flag any row with predictability < 0.6 as "coaching
needed" and name the owner persona responsible.

Expected outcome: ASCII table + coaching flag list + owner persona assignment. Coaching language never names individual contributors.

Prompt 3: Quality-gate drift alert

Goal: detect drift in the 58-gate PHOENIX registry.

Compare the current PI's aggregate quality-gate pass rate against
the 3-PI rolling baseline. For every gate whose pass rate drifts by
more than 10 percentage points, emit a drift row with:
- Gate ID
- Direction (improving / regressing)
- Suspected upstream cause (cite 1 CHANGELOG entry or ADR)
- The owner persona to notify

Publish to docs/safe/portfolio/quality-drift-<pi>.md.

Expected outcome: a Markdown drift report with cited causes and persona ownership rather than raw percentages alone.

Prompt 4: Portfolio heatmap authoring

Goal: build a portfolio-epic heatmap keyed to owner persona.

Read docs/safe/portfolio/epics.md front matter for the nine v1.5.0
portfolio epics. Produce an ASCII heatmap with:
- Rows = epic
- Columns = PI
- Cells = completion bucket (0-25, 26-50, 51-75, 76-100)
- Color legend for each bucket

Then render an identical Streamlit plotly version and link both
from docs/safe/pi-calendar.md.

Expected outcome: dual-format heatmap with an unambiguous legend and a cross-link from the PI calendar.

Prompt 5: Reproducibility manifest

Goal: show how SMC captures metric lineage.

For the scorecard you produced in Prompt 1, author a metrics-run
manifest (JSON) with:
- input artifact URIs + SHA256 for each
- aggregation script path + git SHA
- output artifact path
- OpenTelemetry span IDs emitted
- timestamp (UTC, ISO 8601)

Explain in 150 words why this manifest is the backbone of the
non-surveillance + reproducibility constitutional hard-stops.

Expected outcome: manifest JSON, rationale paragraph tying the manifest to SMC's hard-stop rules.

Prompt 6: Cross-ART vocabulary alignment

Goal: collaborate with SFBI to keep metric names consistent.

You (SMC) want to rename "cycle time" to "lead time" across the
three ARTs to align with Kanban research. Draft the proposal as a
paired artifact:
1. Your metrics-side renaming plan (SMC).
2. The layer-translation hint SFBI should embed in its reconciler
   so SAFe practitioners reading the Business layer and FCC
   practitioners reading the Engineer layer both understand.

Close with a 3-bullet risk list for the rename.

Expected outcome: paired proposal, SFBI translation hint is explicit, risk list covers at least one risk for each ART vocabulary.

Related

• Tutorial: docs/tutorials/advanced-capabilities/performance-optimization-guide.md (Pool B)
• Persona YAML: src/fcc/data/personas/safe_metrics_crafter.yaml
• Decoder: docs/ecosystem/codename-decoder.md
• ADR: docs/decisions/ADR-007_safe_skeleton.md (v1.5.1)
• Other persona prompts: persona-sfbi-prompts.md, persona-tldr-prompts.md