burnwatch 0.7.0 → 0.8.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "burnwatch",
-  "version": "0.7.0",
+  "version": "0.8.0",
   "description": "Passive cost memory for vibe coding — detects paid services, tracks spend, injects budget context into your AI coding sessions.",
   "type": "module",
   "bin": {

package/skills/burnwatch-interview/SKILL.md

@@ -0,0 +1,153 @@
+---
+name: burnwatch-interview
+description: Conversational service interview — the agent walks the user through confirming plans, budgets, and API keys for each detected service. Use after burnwatch init when running in an agent context (non-TTY), or when the user says /burnwatch-interview.
+user-invocable: true
+allowed-tools: Read, Bash, Glob, Grep
+---
+
+# /burnwatch-interview — Agent-Driven Service Interview
+
+You are the interviewer. burnwatch has auto-detected services and applied defaults — your job is to walk the user through confirming or correcting each one in natural conversation.
+
+## Step 1: Get the current state
+
+Run this to get the structured state of all detected services:
+
+```bash
+cd $PROJECT_ROOT && node ./node_modules/.bin/burnwatch interview --json
+```
+
+If burnwatch is installed globally or via npx, use that path instead. Parse the JSON output.
+
+## Step 2: Present the overview
+
+Start with a brief summary:
+
+> I found **N services** in your project. Right now they're configured with defaults — let me walk you through each one so we get accurate tracking.
+>
+> **Current state:** X with API keys (LIVE), Y estimated, Z with no tracking (BLIND)
+> **Total budget:** $NNN/mo (all defaults — let's fix that)
+
+## Step 3: Interview by risk group
+
+Walk through services grouped by risk (the JSON is pre-sorted). For each group:
+
+### LLM / AI Services (ask first — highest variable cost)
+- These are the most important to get right
+- Ask: "What's your actual monthly spend on Anthropic/OpenAI? The default is $100 — too high? too low?"
+- If they have an admin API key, offer to probe: "If you give me your admin key I can check your actual spend"
+
+### Usage-Based Services (Scrapfly, Browserbase, Stripe)
+- Credit-pool services: "What Scrapfly plan are you on? The probe detected Pro with 1M credits — correct?"
+- Ask about actual budget, not just plan cost
+
+### Infrastructure (Vercel, Supabase, AWS)
+- Tiered services: "Are you on Supabase Free or Pro? This matters because Pro is $25/mo with overages"
+
+### Flat-Rate / Free Tier
+- These are usually quick: "PostHog and Inngest are set to Free/$0 — correct? Moving on."
+- Don't dwell on these unless the user has something to say
+
+## Step 4: For each service, use `burnwatch configure`
+
+After the user confirms or corrects, write it back:
+
+```bash
+burnwatch configure --service <id> --plan "<plan name>" --budget <N>
+```
+
+If they provide an API key:
+```bash
+burnwatch configure --service <id> --key "<KEY>" --budget <N>
+```
+
+To exclude a service:
+```bash
+burnwatch configure --service <id> --exclude
+```
+
+The configure command outputs JSON confirming what was written. Check the `success` field.
+
+## Step 5: Show the result
+
+After all services are configured, run:
+
+```bash
+burnwatch status
+```
+
+Present the brief and celebrate what's tracked vs what's still blind.
+
+## Key Behaviors
+
+- **Group questions.** Don't ask 14 questions one at a time. "Your LLMs — Anthropic and OpenAI. What plans and budgets?" is one question, not two.
+- **Lead with what you know.** If the probe detected their Scrapfly plan, say "I detected you're on Pro with 1M credits and you've used 250K so far. Budget $100/mo? [Y/n]" — don't make them scroll through options.
+- **Be brief for obvious services.** Free tier PostHog? "PostHog: Free, $0. Next."
+- **Highlight what matters.** If Anthropic is at $87 of a $100 budget, say so. If Scrapfly credits are 85% consumed, flag it.
+- **Use the probe data.** The interview JSON includes `probeResult` for services where we got live data. Show it: "Your Anthropic spend this month is $47.23 so far."
+- **Respect the user's time.** If they say "defaults are fine for everything" — write them all and move on. Don't force 14 confirmations.
+- **One `configure` call per service.** Don't batch. This makes it clear what was written and lets the user correct individual services.
+
+## Example Flow
+
+```
+Agent: I found 8 services in your project. Let me walk through each one
+       to make sure we're tracking accurately.
+
+       Right now everything has default budgets — let's fix that.
+
+       First, your LLMs:
+       • Anthropic — defaulted to "API Usage" plan, $100/mo budget.
+         I probed your API and you've spent $47.23 this month.
+       • OpenAI — defaulted to "API Usage", $100/mo. No admin key,
+         so I can't check actual spend.
+
+       What are your actual budgets for these?
+
+User:  Anthropic $150 is better, I've been using it heavily.
+       OpenAI keep at $100, and here's my admin key: sk-admin-xxx
+
+Agent: Got it. Let me update both...
+
+       [runs: burnwatch configure --service anthropic --budget 150]
+       [runs: burnwatch configure --service openai --key sk-admin-xxx --budget 100]
+
+       ✅ Anthropic: API Usage | LIVE | $150/mo
+       ✅ OpenAI: API Usage | LIVE | $100/mo (probed: 12.3K tokens this month)
+
+       Next, Scrapfly — I detected you're on Pro ($100/mo, 1M credits).
+       You've used 250K credits so far. Keep the $100 budget?
+
+User:  Yeah that's right.
+
+Agent: [runs: burnwatch configure --service scrapfly --plan pro --budget 100]
+       ✅ Scrapfly: Pro | LIVE | $100/mo | 1M credits
+
+       PostHog, Inngest, and Resend are all on free tiers at $0.
+       Sound right?
+
+User:  Yep.
+
+Agent: [runs configure for each]
+
+       All done! Here's your updated spend brief:
+       [runs: burnwatch status]
+```
+
+## Probe Data Reference
+
+Services with probes and what they return:
+
+| Service | Confidence | What's detected |
+|---------|-----------|-----------------|
+| Scrapfly | high | Plan name, credits used/total |
+| Vercel | high | Plan tier (hobby/pro/enterprise) |
+| Supabase | high | Plan tier (free/pro/team) |
+| Anthropic | medium | USD spend this billing period |
+| OpenAI | medium | Token usage this period |
+| Stripe | medium | Balance (available + pending) |
+| Browserbase | medium | Session count, browser hours |
+| Upstash | low | Database count (key validation) |
+| PostHog | low | Org found (key validation) |
+| Gemini, Resend, Inngest, Voyage AI | none | No API — ask user directly |
+| AWS | none | Too complex — flag as BLIND |