competition-analysislisted

# Competition Analysis **This skill computes nothing itself.** Every number and label below is already computed by deterministic, tested Python in the wrapped engine — `hunt deepen-start` runs the competitor autopsy and lineage checks and prints the result. Your job is only to read that result correctly and interpret it using the exact thresholds in `references/competition.md` — never recompute a concentration figure, never round differently, never relabel a level the engine already assigned. ## When to use this After a candidate has reached `ECONOMICS_CHECK` (via `amazon-validation` and economics being set). This is the entry point to the "deepen" pipeline: `COMPETITOR_ANALYSIS → VOC_ANALYSIS → DIFFERENTIATION_ANALYSIS → RED_TEAM → HUMAN_REVIEW`. ## Steps 1. **Run the autopsy.** ``` hunt deepen-start <run_id> ``` This is deterministic, no LLM involved. It advances every eligible `ECONOMICS_CHECK` opportunity in the run to `COMPETITOR_ANALYSIS` and prints a JSON `competitor_result` per opportunity. Note every printed `opportunity_id` — each needs its own pass through the rest of this pipeline. 2. **Read the concentration figures, not just the labels.** For each of `review_concentration`, `brand_concentration`, `oos_frequency`, check: - The reported value (or `None`/`UNKNOWN`). - `competitor_set_size` (post parent-ASIN-dedup — this is the number that matters) vs `competitor_set_size_before_parent_dedup` (raw rows). If these diff