@witchpot/gmc 0.1.7 → 0.1.8

package/CHANGELOG.md

@@ -2,6 +2,14 @@
 
 All notable changes to the Game Market Copilot CLI will be recorded here.
 
+## 0.1.8 - 2026-06-12
+
+- Updated the bundled `gmc-analysis` skill to v0.2.0 with the learnings from
+  the first price-band field test: camelCase-query to kebab-case-flag
+  mapping note, zero-review rating artifact pitfall, discounted snapshot
+  price pitfall, price-question routing to explicit-bucket aggregates, and
+  the evidence-pack quota location exception.
+
 ## 0.1.7 - 2026-06-12
 
 - Bundled the `gmc-analysis` Agent Skill (v0.1.0): analysis recipes, pivot

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@witchpot/gmc",
-  "version": "0.1.7",
+  "version": "0.1.8",
   "private": false,
   "description": "Command line tools for Game Market Copilot market research.",
   "license": "UNLICENSED",

package/skill/gmc-analysis/CHANGELOG.md

@@ -1,5 +1,20 @@
 # gmc-analysis skill changelog
 
+## 0.2.0 - 2026-06-12
+
+- Folded in the five learnings from the first field test (2025 city
+  builders, price-band vs rating/review analysis; 7/7 rubric pass):
+  - setup: documented the camelCase-query -> kebab-case-flag mapping and
+    that per-command `--help` is currently generic;
+  - new pitfall P5: zero-review games aggregate as rating 0 — pair rating
+    metrics with `--reviews-min`, derive per-bucket hit rates from counts;
+  - new pitfall P6: snapshot prices include active discounts — check the
+    `discount` field before price-band conclusions;
+  - recipe R4: price/distribution questions route to aggregate with
+    explicit buckets first; `cohort pricing` is page-bounded texture only;
+  - quota: composed evidence packs carry quota under
+    `meta.upstreamRequests[].entitlements`, not at the top level.
+
 ## 0.1.0 - 2026-06-12
 
 - Initial release, seeded from the 2026-06-12 field test (roguelike cohort

package/skill/gmc-analysis/SKILL.md

@@ -7,7 +7,7 @@ description: >-
   about Steam games, reviews, creators, showcases, or campaigns that gmc can
   answer.
 metadata:
-  version: 0.1.0
+  version: 0.2.0
 ---
 
 # GMC Analysis
@@ -26,8 +26,11 @@ gmc auth status --json        # anonymous => redacted previews + tiny quota
 gmc schema --json             # the authoritative command/endpoint reference
 ```
 
-Never guess flags: `gmc schema --json` and `gmc <command> --help` are the
-source of truth for commands and parameters. Always pass `--json`.
+Never guess flags: `gmc schema --json` is the source of truth for commands
+and parameters. Endpoint query params are camelCase and the matching CLI
+flag is the kebab-case equivalent (`priceMin` -> `--price-min`,
+`reviewsGt` -> `--reviews-gt`). Per-command `--help` currently prints one
+generic usage list, so trust the schema. Always pass `--json`.
 
 ## Core workflow for cohort questions
 

package/skill/gmc-analysis/references/pitfalls.md

@@ -42,7 +42,29 @@ the quota is tiny, which can read as "the API has no data".
 **Rule**: run `gmc auth status --json` first; if `authMode` is `anonymous`,
 say so in your answer and scope expectations to public preview data.
 
-## P5. Don't paginate by brute force
+## P5. Zero-review games carry rating 0 in aggregates (2026-06)
+
+Unconditional `avg_rating` / `median_rating` in low-traction segments (cheap
+or free price buckets) is crushed by zero-review games aggregating as
+rating 0 (observed: a free bucket with avg_rating 11.1). This misreads as
+"cheap games are hated" when it actually means "most cheap games have no
+reviews".
+
+**Pivot**: pair rating metrics with `--reviews-min` (e.g. 10). For "share of
+well-rated games per bucket", run `--metric count` twice over the same
+bucket spec — once unconditional, once with `--reviews-min 10
+--rating-min 80` — and divide.
+
+## P6. Snapshot prices include active discounts (2026-06)
+
+`price` is the current store price, not the launch/list price; a discounted
+title lands in a lower price bucket (observed: a $34.99 list price counted
+as $23.44 at -33%). Game rows carry a `discount` field.
+
+**Rule**: when price-band conclusions matter, check `discount` on the
+exemplar rows and caveat bucket assignments, especially during sale seasons.
+
+## P7. Don't paginate by brute force
 
 List endpoints cap at `limit=100`. For "how many" questions use
 `gmc games count`; for distributions use `gmc games aggregate` — both are

package/skill/gmc-analysis/references/quota.md

@@ -9,7 +9,10 @@ indicative — the response metadata is the source of truth.
 
 1. **Read before you spend.** Every bearer response carries
    `meta.entitlements.quota` = `{ used, limit }`. Check it at the start of a
-   session and after each fan-out.
+   session and after each fan-out. Exception: CLI-composed evidence packs
+   (`cohort ...`, `reviews patterns`, `games benchmark`, ...) expose quota
+   per upstream call instead — read the last entry of
+   `meta.upstreamRequests[].entitlements.quota`.
 2. **Estimate fan-out cost first.** A cohort sweep costs roughly
    `search pages + (1 per game per section call)`. A 114-game cohort with one
    analysis call each ~= 120 requests. Compare against remaining quota and

package/skill/gmc-analysis/references/recipes.md

@@ -80,6 +80,19 @@ gmc games aggregate --source steam --metric count --group-by coverage <filters>
 `data.basis = server_aggregate` means market-wide for the filter set (within
 GMC source coverage) — unlike evidence packs, these are not page-bounded.
 
+**Price/distribution questions belong here, not in `cohort pricing`.** Run
+the aggregate with explicit buckets first (e.g.
+`--bucket price:0,5,10,15,20,25,30,40,60`); `gmc cohort pricing` is
+returned-page (top ~20) texture, useful only as a supplement and must be
+labeled as page-bounded evidence.
+
+Per-bucket "hit rate" pattern (one request per condition): pair the same
+bucket spec with shared filters — `--reviews-min 500 --metric count` gives
+reach per price band; `--reviews-min 10 --rating-min 80 --metric count`
+gives the well-rated share. Divide by the unconditional bucket counts.
+Never read unconditional rating aggregates in low-traction buckets
+(pitfall P5).
+
 ## Output contract for every analysis
 
 Always include in your final answer: the cohort definition and size, how many

package/src/constants.js

@@ -8,7 +8,7 @@ export const EXIT_API = 4;
 
 export const CLI_PACKAGE_NAME = "@witchpot/gmc";
 export const CLI_BETA_PACKAGE_NAME = "@witchpot/gmc";
-export const CLI_VERSION = "0.1.7";
+export const CLI_VERSION = "0.1.8";
 
 export const DEFAULT_API_BASE_URL = "https://gamemarketcopilot.com";
 export const DEFAULT_CONFIG_PATH = path.join(homedir(), ".config", "gmc", "config.json");