cost-modellisted

# Cost model — make great-pm cost claims defensible great-pm reports cost numbers on the board (via /pm-cost, /pm-burn, /pm-digest). Those numbers MUST be auditable. The classic mistake: a shiny "1000× human equivalent" claim that doesn't survive scrutiny. This skill defines the format and the discipline. ## The 4-line cost section Every great-pm initiative's cost section follows this exact template: ```markdown ## Cost estimate **LLM**: $<low>–<high> (<N> agent invocations × $<per-call avg>) **Human equiv**: $<low>–<high> (<hours> × $<rate>/h) **Infra delta**: $<low>–<high>/month **Time to ship**: <hours> agent-time, <hours/days> wall-clock > Methodology: <one-sentence rationale for each range> ``` ### Why this exact format? Board parsers anchor on **line-start** "LLM" and "Human" labels. Mid-line references are ignored. Stick to the template — drift breaks the board. ## How to estimate each line ### LLM cost For each great-pm agent that runs in the initiative, estimate: - **Prompt tokens** = system prompt + injected context (PROJECT.md + brain.md tail + HANDOFF.md, per SubagentStart hook) + agent-specific input - **Completion tokens** = typical output for that agent type Quick reference for the **`opus` alias** — great-pm's default model, which always resolves to the newest Opus (Opus 4.8 at time of writing): **$5/M input, $25/M output** (verified 2026-07-21 against Anthropic's price sheet — re-verify at each model release; the alias tracks new Opus versions a