surpluslisted

# Surplus Burn Mode — Phase 1 (Scan + Rank) > **Read-only.** Phase 1 does **NO execution** and **NO mutation** of any backlog source. > It writes **exactly two artifacts**: the ranked plan (`.md`) and its machine-readable > sidecar (`.json`). Nothing else on disk changes. ## What this mode does Scans TAD's backlog sources + one OBJECTIVES-driven generator, ranks every candidate **value-first** (expected_value = value × confidence; density is only an efficiency tiebreaker), tags risk/safety, and produces a ranked plan you can act on to consume unused weekly Claude usage on the highest-value work. It does **not** pick anything up and run it. Choosing what to execute (and actually executing it) is **Phase 2** — see below. ## Commands ### `*surplus --plan` (and bare `*surplus`) 1. **Stamp the date.** Compute today's date string `YYYY-MM-DD` (the date lives at this SKILL boundary so the workflow stays free of `Date.now()` / `new Date()`). Call it `<stamp>`. Use the **same** `<stamp>` for the filename and the artifact header. 2. **Invoke the scan workflow:** ``` Workflow({ name: 'surplus-scan', args: { date: '<stamp>', output_path: '.tad/active/SURPLUS-PLAN-<stamp>.md' } }) ``` 3. **Persist the two artifacts.** The workflow runtime is sandboxed (no filesystem access), so it returns the rendered content. Write them yourself with the Write tool: - `result.plan_markdown` → `result.plan_path` (`.tad/active/SURPLUS-PLAN-<stam