🧬 Prompt Augmentation — the initialization prompt each persona is built with at session start
Timesheet — real Claude usage & API-equivalent cost (stardev.timesheet · PG)
Engineer
Station
Tier
Tickets
Active hrs
Tokens
Work-value ($ proxy)
Claude Code consumption & seat licenses
Engineer
Station
Tier
Tokens (7d)
Cost 7d (API-eq)
Cost 30d (proj)
Real per-seat usage from archived Claude sessions, priced at per-model API rates (cache priced separately). Engineers run on flat-rate subscriptions, so the actual bill is far lower — this shows the value/throughput delivered, not the invoice.
Token utilization by type — per persona & per model (stardev.session_tokens · PG)
Fed every 2h from each devstation's own Claude Code transcripts; the model is the REAL per-call model, not a fixed tier. Pick a window:
Persona
Engineer
Model(s) used
Input
Output
Cache-create
Cache-read
Total
Cache reads dominate the totals (the prompt cache replayed each turn). Any $ figure derived from these is a notional per-token-equivalent, NOT the actual bill — engineers run on flat-rate Max subscriptions.
Devstations & personas — live resources & best-to-assign
Live CPU/mem + ticket load per station; ranks who is most available for new work. Refreshes every 30s.
⚙️ Runner jobs — live CI / GitHub Actions status
By workflow
Workflow
✓ success
✗ failure
● in-prog
queued
cancelled
Recent runs (last 60)
When
Workflow
Branch
Status
By
Star Developers — App Builder (prototype). Compliance reflects the live filesystem (13-doc suite: 7 Formation + 6 excellence-layer blueprints); engineer jobs from engineering.tickets; timesheet persisted to stardev.timesheet (PG). Auto-regenerated by the delivery-manager harvest loop.