fix onboard

linkedin-lead-filter/SKILL.md

@@ -77,7 +77,8 @@ C) I don't have a CSV — show me how signals work
 **If user picks A or provides a file directly:**
 Read the file they provide. If pasted inline, parse it as CSV.
 
-Acknowledge the leads briefly, then show the signal showcase before proceeding:
+Acknowledge the leads briefly, then show the signal showcase before proceeding.
+(Only show the btw block for option A — skip it for option C.)
 
 ```
 Got it — [N] leads loaded from [filename or "your CSV"].
@@ -136,9 +137,21 @@ Say:
 Then stop. Do not proceed until they have a file.
 
 **If user picks C:**
-Read `sample_warm_leads.example.csv` from the repo root and proceed with that as the input. Tell the user:
+Read `sample_warm_leads.example.csv` from the repo root. Tell the user:
 > Loading sample warm leads — placeholder contacts with realistic signals so you can run the full pipeline right now. Swap in your real leads any time.
 
+Do **not** show the btw/signals table.
+
+Display the sample leads as a readable table (Name, Title, Company, Signal columns). Then ask:
+
+```
+Here are your sample leads. Want to change anything before I run the filter?
+(edit a row, swap someone out, add a lead — or just say "looks good")
+```
+
+If they request changes, apply them, re-display the updated table, and ask again.
+Once they confirm ("looks good" or equivalent), proceed with the (possibly edited) leads as the input.
+
 Identify the columns available. At minimum, look for: Name, Title, Company, LinkedIn URL, and any signal or engagement data.
 
 If key columns are missing or ambiguous, do not guess. Ask for column mapping using:

onboard/SKILL.md

@@ -133,39 +133,28 @@ If C: ask for each ICP field individually, then ask for competitor description a
 
 ## Step 4: Collect conversion goal
 
-Ask:
+The user already provided a landing page URL in Step 1. Use it as the default. Ask:
-```
-What's the goal of your outreach — what action do you want an interested prospect to take?
 
-A) Visit a landing page
+```
-B) Book a call
+What do you want an interested prospect to do?
-C) Fill out a form
+
-D) Receive a lead magnet (e.g. guide, case study)
+A) Visit your landing page — [URL from Step 1]
-E) Subscribe to a newsletter
+B) Book a call — paste a Calendly or booking link
+C) Fill out a form — paste the form URL
+D) Receive a lead magnet — paste the link
+E) Subscribe to a newsletter — paste the link
 F) Other — describe it
 
-If you have a link for it, paste it now. You can also skip this and add it later.
 ```
 
-Save the choice as `outreach.conversionGoal.type` and the link (if provided) as `outreach.conversionGoal.link`. If no link is given, leave `link` as `null`. Do not block progress if the link is missing.
+Wait for the user to reply before continuing. Do not assume A.
+
+If user confirms A: set `outreach.conversionGoal.type` to `landing_page` and `outreach.conversionGoal.link` to the URL from Step 1.
+Otherwise: save their choice as type and their link as the URL. If no link given, leave `link` as `null`. Do not block progress.
 
 ---
 
-## Step 5: Collect timezone and cadence
+## Step 5: Save settings
-
-Ask:
-```
-Two quick preferences:
-
-1. Your timezone? (e.g. America/New_York, Europe/London, Asia/Singapore)
-2. How often do you want to run outreach?
-   A) Daily — pick a time (e.g. 08:00)
-   B) Weekly — pick a day + time (e.g. Monday 08:00)
-```
-
----
-
-## Step 6: Save settings
 
 Write the confirmed ICP + preferences into `state/linkedin-settings.json`:
 
@@ -175,8 +164,6 @@ Write the confirmed ICP + preferences into `state/linkedin-settings.json`:
 - `targetMarket.geographies`
 - `targetMarket.alwaysExcludeRoles`
 - `qualification.disqualifyCategories` — array of snake_case labels derived in Step 2 (e.g. `["competitors", "non_decision_makers", "b2c_focused"]`)
-- `timezone`
-- `cadence`
 - `outreach.conversionGoal.type`
 - `outreach.conversionGoal.link` (may be `null`)
 - `firstRunCompleted` → `true`
@@ -197,55 +184,7 @@ Also write the competitor profile to `state/competitor-profile.json`:
 
 ---
 
-## Step 6.5: Telemetry opt-in (one time only)
+## Step 6: Confirm and hand off
-
-Check if `state/linkedin-system-state.json` has `"telemetryPrompted": true`. If yes, skip this step entirely.
-
-Ask once using AskUserQuestion:
-
-> Help us build better signals for your industry! Community mode shares anonymous usage data (which skills you use, how many leads processed, signal types that worked) — no names, no message content, no file paths ever sent.
-> Change anytime by setting `"telemetry": "off"` in `state/linkedin-settings.json`.
-
-Options:
-- A) Sure, happy to help improve this
-- B) No thanks
-
-If A: set `settings.telemetry = "community"` in `state/linkedin-settings.json`
-If B: ask one follow-up:
-
-> How about anonymous mode? We just learn that *someone* ran the pipeline — no ID, no details. Just a counter that helps us know the tool is useful.
-
-Options:
-- A) Anonymous is fine
-- B) No thanks, fully off
-
-If B→A: set `settings.telemetry = "anonymous"`
-If B→B: set `settings.telemetry = "off"`
-
-Always set `state.telemetryPrompted = true` in `state/linkedin-system-state.json` after this step regardless of choice.
-
-**What gets sent (community mode only):**
-- Which skill was run
-- Number of leads processed
-- Signal types present (e.g. `post_like`, `funding_news`) — no names or content
-- Outcome (done / blocked / needs_context)
-- Timestamp
-
-**What is never sent:** names, companies, LinkedIn URLs, message content, file paths, your landing page URL.
-
-**How it's sent:** at the end of each skill run via:
-```bash
-curl -s -X POST https://linkednav-telemetry.vercel.app/api/telemetry \
-  -H "Content-Type: application/json" \
-  -d '{"skill":"<skill_name>","leads":<count>,"signals":[<types>],"outcome":"<outcome>","ts":"<iso>"}' \
-  2>/dev/null &
-```
-
-Only run this curl if `settings.telemetry` is `"community"`. If `"anonymous"`, send `{"skill":"<skill_name>","ts":"<iso>"}` only. If `"off"`, skip entirely.
-
----
-
-## Step 7: Confirm and hand off
 
 Output:
 
@@ -257,8 +196,6 @@ Industries:   [list]
 Seniority:    [list]
 Company size: [list]
 Geographies:  [list]
-Timezone:     [tz]
-Cadence:      [frequency + time]
 Goal:         [conversionGoal.type] — [conversionGoal.link or "no link set"]
 
 Settings saved to: state/linkedin-settings.json

scripts/bootstrap-system.js

@@ -1,6 +1,7 @@
 #!/usr/bin/env node
 const fs = require("fs");
 const path = require("path");
+const readline = require("readline");
 const { readJson, writeJson, ensureDir } = require("./lib/io");
 
 const ROOT = path.resolve(__dirname, "..");
@@ -62,9 +63,60 @@ if (!fs.existsSync(brandVoicePath)) {
   );
 }
 
-console.log(JSON.stringify({
+const result = {
   ok: true,
   settingsInitialized: !currentSettings,
   stateInitialized: !currentState,
   root: ROOT
-}, null, 2));
+};
+
+// Telemetry opt-in — ask once on first setup
+const freshState = readJson(statePath, {});
+if (!freshState.telemetryPrompted) {
+  const rl = readline.createInterface({ input: process.stdin, output: process.stdout });
+
+  const ask = (q) => new Promise((resolve) => rl.question(q, resolve));
+
+  (async () => {
+    console.log("\n──────────────────────────────────────────");
+    console.log("Help us improve these skills for your industry.");
+    console.log("Community mode shares anonymous usage data — which skills you run,");
+    console.log("how many leads processed, signal types that worked.");
+    console.log("No names, no message content, no file paths. Ever.");
+    console.log("Change anytime: set \"telemetry\": \"off\" in state/linkedin-settings.json");
+    console.log("──────────────────────────────────────────");
+
+    const a1 = await ask("\nA) Sure, happy to help improve this\nB) No thanks\n> ");
+
+    let telemetryLevel;
+
+    if (a1.trim().toUpperCase() === "A") {
+      telemetryLevel = "community";
+    } else {
+      const a2 = await ask(
+        "\nHow about anonymous mode? We just learn that someone ran the pipeline —\n" +
+        "no ID, no details. Just a counter that helps us know the tool is useful.\n\n" +
+        "A) Anonymous is fine\nB) No thanks, fully off\n> "
+      );
+      telemetryLevel = a2.trim().toUpperCase() === "A" ? "anonymous" : "off";
+    }
+
+    rl.close();
+
+    // Save telemetry choice to settings
+    const settings = readJson(settingsPath, {});
+    if (!settings.settings) settings.settings = {};
+    settings.settings.telemetry = telemetryLevel;
+    writeJson(settingsPath, settings);
+
+    // Mark telemetry as prompted in state
+    const state = readJson(statePath, {});
+    state.telemetryPrompted = true;
+    writeJson(statePath, state);
+
+    result.telemetry = telemetryLevel;
+    console.log("\n" + JSON.stringify(result, null, 2));
+  })();
+} else {
+  console.log(JSON.stringify(result, null, 2));
+}

scripts/run-lead-filter.js

@@ -103,6 +103,7 @@ function disqualifyCategory(record, title, company, industry, about) {
   const c = lower(company);
   const i = lower(industry);
   const a = lower(about);
+  const activeCategories = new Set((settings.qualification.disqualifyCategories || []).map((v) => lower(v)));
   const roleExcludes = (settings.targetMarket.alwaysExcludeRoles || []).map((v) => lower(v));
   const companyExcludes = (settings.targetMarket.alwaysExcludeCompanyTypes || []).map((v) => lower(v));
   const competitorCompanyKw = (competitorProfile ? competitorProfile.companyKeywords || [] : []).map((v) => lower(v));
@@ -110,26 +111,28 @@ function disqualifyCategory(record, title, company, industry, about) {
   const competitorServiceKw = (competitorProfile ? competitorProfile.serviceKeywords || [] : []).map((v) => lower(v));
   const allCompetitorCompanyKw = [...new Set([...companyExcludes, ...competitorCompanyKw])];
 
-  if (
+  if (activeCategories.has("non_decision_makers") && (
     /student|intern|assistant|junior/.test(t) ||
     roleExcludes.some((r) => r && t.includes(r))
-  ) {
+  )) {
     return "non_decision_makers";
   }
 
-  if (
+  if (activeCategories.has("competitors") && (
     allCompetitorCompanyKw.some((k) => k && c.includes(k)) ||
     competitorIndustryKw.some((k) => k && i.includes(k)) ||
     competitorServiceKw.some((k) => k && a.includes(k))
-  ) {
+  )) {
     return "competitors";
   }
 
-  if (/consumer|retail|b2c/.test(i) && !/b2b|software|saas|services/.test(i)) {
+  if (activeCategories.has("b2c_focused") &&
+    /consumer|retail|b2c/.test(i) && !/b2b|software|saas|services/.test(i)) {
     return "b2c_focused";
   }
 
-  if (/engineer|developer|support|it admin|back office/.test(t) && !/head|director|vp|chief/.test(t)) {
+  if (activeCategories.has("no_outbound_intent") &&
+    /engineer|developer|support|it admin|back office/.test(t) && !/head|director|vp|chief/.test(t)) {
     return "no_outbound_intent";
   }
 

state/linkedin-settings.example.json

@@ -1,12 +1,6 @@
 {
   "version": "1.0.0",
   "firstRunCompleted": false,
-  "timezone": "America/Los_Angeles",
-  "cadence": {
-    "frequency": "daily",
-    "dayOfWeek": "monday",
-    "time": "08:00"
-  },
   "targetMarket": {
     "industries": [],
     "companySizes": [],