cost-modellisted

# cost-model ## Purpose Synthesize existing cost state into forward-looking projections so an operator can answer "what will Glam burn this month?" or "is the LLM bill on track?" without inspecting raw JSON. cost-model is read-only — it never mutates `quota_state.json` or `costs.json`. It produces a markdown report plus a machine-readable JSON sibling for downstream skills (notably cost-guard, F002). Use this skill when an operator needs a current cost picture or before making a spend decision (e.g., approving a tier promotion, sizing a new agent panel). ## Arguments | Argument | Required | Description | |---|---|---| | `--costs` | no | Path to costs.json (default: `dashboard/state/costs.json`) | | `--quota` | no | Path to quota_state.json (default: `~/.jarvis/quota_state.json`) | | `--revenue` | no | Path to revenue.json (default: `dashboard/state/revenue.json`) — optional input; absence is fine | | `--out` | no | Output directory for the report pair (default: `evidence/cost-model/`) | | `--as-of` | no | Override "now" with an ISO-8601 timestamp for deterministic golden-output tests | | `--fixtures` | no | Directory containing `costs.json`, `quota_state.json`, and (optional) `revenue.json` — convenient for testing | If no arguments: reads the default paths and writes to `evidence/cost-model/`. ## Prerequisites - `scripts/quota_check.py` (F020) has been run at least once so `quota_state.json` exists. - `scripts/maintainer/refresh-costs.sh` has been run at least once s