Update to Fullstory

Changed files

Choose a file to see its text changes. Binary or large files show metadata only.

.app.json →

.codex-plugin/plugin.json →

.mcp.json →

assets/logo.png →

skills/comparisons/SKILL.md →

skills/friction-analysis/SKILL.md →

skills/general-analysis/SKILL.md →

skills/general-analysis/references/funnels.md →

skills/general-analysis/references/journeys.md →

skills/general-analysis/references/metrics.md →

skills/general-analysis/references/sessions.md →

skills/general-analysis/references/validation.md →

skills/session-review/SKILL.md →

.codex-plugin/plugin.json

− removed · + added

--- before
+++ after
@@ -1,27 +1,30 @@
 {
-  "apps": "./.app.json",
   "author": {
     "name": "Fullstory"
   },
-  "description": "Ask about your product analytics, and get answers you can verify.\n\nFullstory MCP connects Codex and ChatGPT to your product's behavioral data. Ask how many users adopted a feature, where checkout is leaking, or whether yesterday's release broke something, and Codex and ChatGPT build the segment, computes the metric, and explains the result with a link back to Fullstory.\n\nAnalytics connectors can give you numbers. Fullstory shows you what happened. Metrics unpack into real session replay, so \"conversion dropped 12%\" becomes \"the coupon field errors for logged-out users.\"\n\nAnd because Fullstory autocaptures clicks, page views, errors, and frustration signals, many questions need no instrumentation at all. Ask about rage clicks on pricing or errors in checkout without shipping a single event.\n\nWhat Codex and ChatGPT can do:\n(1) Build and save segments and metrics.\nDescribe the users or behavior you care about in your own words; Codex and ChatGPT create segments or metrics in Fullstory where your whole team can reuse it.\n\n(2) Compute on demand.\n Counts, rates, trends, and breakdowns by device, browser, geography, or user property, over any time window.\n\n(3) Speak your product's language. \nCodex and ChatGPT discover your pages, named elements, and defined events, so questions map to your real product vocabulary instead of guessed selectors.\n\n(4) Analyze conversion.\nRead your saved funnels and measure where users drop and why.\n\n(5) Back every number with sessions.\nRetrieve the sessions and events behind any metric, including frustration signals like rage clicks and dead clicks.\n\n(6) Read sessions at scale.\nAsk Codex and ChatGPT to review dozens of sessions behind a drop-off or an alert and report the patterns, including failures that never reached your error tracker.\n\nTeams use it to let PMs, support, and engineers self-serve answers that used to queue behind an analyst, and to settle \"why did this number move?\" with evidence instead of theories.",
+  "description": "See what actually happened behind every number.\n\nFullstory MCP gives ChatGPT and Codex both halves of product investigation: analytics to size a problem and session replay to see why it happened. Ask where checkout is leaking, why a metric moved, or how to reproduce a bug. It discovers your pages and elements, computes the metric or funnel, and opens the sessions behind the drop. Every answer links to Fullstory, so you can check the work in one click.\n\nFullstory shows what users experienced - including silent failures that never threw an error - so \"conversion dropped 12%\" becomes \"the coupon field fails 5x more for logged-out users.\"\n\nBecause every answer links to the proof, you can give ChatGPT or Codex a standing job. Set up a scheduled task once, and it reports back with the sessions attached: an hourly watch for silent failures, automatic bug scoping on new tickets, or a Monday ranking of the fixes that would recover the most users.\n\nWhat ChatGPT and Codex can do:\n\n(1) Map a session, then zoom to the moment. Read a compact timeline first, then jump to the timestamp where it broke to pull the screenshot and page DOM from before and after - keeping context light.\n\n(2) Review sessions at scale. Because sessions load as compact structured data, read dozens of replays in one run and surface why users gave up.\n\n(3) Build segments, metrics, and funnels in your own words. Discover your pages, named elements, and events so questions map to your product.\n\n(4) Compute on demand. Run counts, rates, trends, funnels, journeys, and friction breakdowns by device, browser, geography, or user property.\n\nFullstory autocaptures clicks, page views, console and network errors, and frustration signals (rage and dead clicks), so many questions need no instrumentation at all.",
   "interface": {
     "capabilities": [],
     "category": "Data & Analytics",
     "defaultPrompt": [
-      "Find the page with the highest rage-click rate in the past 7 days. Pull 3 sessions and walk me through what happened.",
-      "Build a segment of users who hit a console error this week, then pull two of their sessions and summarize what happened.",
-      "Which users were highly active 60 days ago but haven't logged in for 30 days? Pull 5 sessions for their last week of activity."
+      "Every hour, find checkout sessions with frustration but no logged error, and flag anything new with a session link.",
+      "Find a session with a console error this week. Show me the screen where it broke, then count who else hit it.",
+      "Find the page with the most rage clicks this week. Open 3 sessions and show me the screen where users got stuck."
     ],
     "developerName": "Fullstory",
     "displayName": "Fullstory",
-    "longDescription": "Ask about your product analytics, and get answers you can verify.\n\nFullstory MCP connects Codex and ChatGPT to your product's behavioral data. Ask how many users adopted a feature, where checkout is leaking, or whether yesterday's release broke something, and Codex and ChatGPT build the segment, computes the metric, and explains the result with a link back to Fullstory.\n\nAnalytics connectors can give you numbers. Fullstory shows you what happened. Metrics unpack into real session replay, so \"conversion dropped 12%\" becomes \"the coupon field errors for logged-out users.\"\n\nAnd because Fullstory autocaptures clicks, page views, errors, and frustration signals, many questions need no instrumentation at all. Ask about rage clicks on pricing or errors in checkout without shipping a single event.\n\nWhat Codex and ChatGPT can do:\n(1) Build and save segments and metrics.\nDescribe the users or behavior you care about in your own words; Codex and ChatGPT create segments or metrics in Fullstory where your whole team can reuse it.\n\n(2) Compute on demand.\n Counts, rates, trends, and breakdowns by device, browser, geography, or user property, over any time window.\n\n(3) Speak your product's language. \nCodex and ChatGPT discover your pages, named elements, and defined events, so questions map to your real product vocabulary instead of guessed selectors.\n\n(4) Analyze conversion.\nRead your saved funnels and measure where users drop and why.\n\n(5) Back every number with sessions.\nRetrieve the sessions and events behind any metric, including frustration signals like rage clicks and dead clicks.\n\n(6) Read sessions at scale.\nAsk Codex and ChatGPT to review dozens of sessions behind a drop-off or an alert and report the patterns, including failures that never reached your error tracker.\n\nTeams use it to let PMs, support, and engineers self-serve answers that used to queue behind an analyst, and to settle \"why did this number move?\" with evidence instead of theories.",
+    "longDescription": "See what actually happened behind every number.\n\nFullstory MCP gives ChatGPT and Codex both halves of product investigation: analytics to size a problem and session replay to see why it happened. Ask where checkout is leaking, why a metric moved, or how to reproduce a bug. It discovers your pages and elements, computes the metric or funnel, and opens the sessions behind the drop. Every answer links to Fullstory, so you can check the work in one click.\n\nFullstory shows what users experienced - including silent failures that never threw an error - so \"conversion dropped 12%\" becomes \"the coupon field fails 5x more for logged-out users.\"\n\nBecause every answer links to the proof, you can give ChatGPT or Codex a standing job. Set up a scheduled task once, and it reports back with the sessions attached: an hourly watch for silent failures, automatic bug scoping on new tickets, or a Monday ranking of the fixes that would recover the most users.\n\nWhat ChatGPT and Codex can do:\n\n(1) Map a session, then zoom to the moment. Read a compact timeline first, then jump to the timestamp where it broke to pull the screenshot and page DOM from before and after - keeping context light.\n\n(2) Review sessions at scale. Because sessions load as compact structured data, read dozens of replays in one run and surface why users gave up.\n\n(3) Build segments, metrics, and funnels in your own words. Discover your pages, named elements, and events so questions map to your product.\n\n(4) Compute on demand. Run counts, rates, trends, funnels, journeys, and friction breakdowns by device, browser, geography, or user property.\n\nFullstory autocaptures clicks, page views, console and network errors, and frustration signals (rage and dead clicks), so many questions need no instrumentation at all.",
     "privacyPolicyURL": "https://www.fullstory.com/legal/privacy-policy/",
     "shortDescription": "Product analytics for agents.",
     "supportURL": "https://www.fullstory.com/contact/",
     "termsOfServiceURL": "https://www.fullstory.com/legal/terms-and-conditions/",
-    "websiteURL": "https://www.fullstory.com/"
+    "websiteURL": "https://www.fullstory.com/",
+    "logo": "./assets/logo.png",
+    "composerIcon": "./assets/logo.png"
   },
   "name": "app-6a9a1ebb11fc81919ef2aa7231b9b068",
   "skills": "./skills",
-  "version": "3.0.0"
-}
\ No newline at end of file
+  "mcpServers": "./.mcp.json",
+  "version": "3.1.0",
+  "apps": "./.app.json"
+}