galiana

Galiana — Discipline Analytics

Step 1: Run analyzer

GALIANA_SCRIPT=$(find ~/.claude/plugins/cache -path '*/clavain/*/galiana/analyze.py' 2>/dev/null | head -1)
[[ -z "$GALIANA_SCRIPT" ]] && GALIANA_SCRIPT=$(find ~/projects -path '*/os/clavain/galiana/analyze.py' 2>/dev/null | head -1)

If found: python3 "$(dirname "$(dirname "$GALIANA_SCRIPT")")/galiana/analyze.py"

If not found: explain Galiana is not installed and stop.

Step 2: Read KPI cache

Read ~/.clavain/galiana-kpis.json. If missing or no KPI payload, show a getting-started message explaining KPI events come from disciplined review/testing workflows and defect logging.

Step 3: Present results

Galiana — Discipline Analytics
════════════════════════════════
Period: YYYY-MM-DD to YYYY-MM-DD
Beads shipped: N | Events analyzed: N

KPIs (diagnostic-first)
────────────────────────
[worst metric first, healthiest last]
Format: KPI Name: value (numerator/denominator)

Agent Scorecard
────────────────
Agent             Findings  P0  P1
fd-architecture   5         0   2
fd-quality        3         1   1

Advisories
──────────
- [info] ...
- [warn] ...

Topology Experiment Results

If topology_efficiency.available is true:

Topology Efficiency (recall vs production)
──────────────────────────────────────────
Task Type      T2    T4    T6    T8    Samples
code_review    0.60  0.85  0.95  0.95  12

Highlight the smallest topology achieving >=90% recall ("sweet spot"). If < 20 total experiments: "Data still accumulating."

Eval Harness Health

If eval_health.available is true:

Eval Harness Health
──────────────────────────────────────────
Overall: 85% property pass rate | Avg recall: 0.92 | 20 total runs

Fixture                  Pass Rate  Recall  Runs
synth-sql-injection      1.00       0.95    4

Highlight any fixture with pass_rate < 1.0. If < 10 total runs: "Limited data."

Offer drill-down

After report: Want per-bead analytics? Run Galiana with --bead <id>.