feat(leads): alert on intent, and match on what you sell

[ralyodio]

Adds prd/0001-intent-lead-monitoring.md — a read of Leadmatically ($19–$1,499/mo) against what we shipped this week — and implements its two P0s.

What they have that we don't

Their scoring is comparable to ours, and in two respects ours is stricter: we refuse people who wrote "no vendors", and we refuse drafts stating anything the campaign never claimed. They beat us on everything that happens after a signal is found — which is where the value of finding it goes.

	them	us (before)
Instant alerts (Email/Slack/Telegram)	✅	❌
Score against a business description	✅	keywords only
Saved reply prompt + drafted reply	✅	Reddit only, not configurable
Outcome tracking per conversation	✅	column exists, nothing writes it
Per-platform analytics	✅	❌

R1 — we were finding requests and telling nobody

A conversation worth answering has a useful life measured in hours. A signal sitting in a table until someone opens the Leads page is indistinguishable from one never found.

• Batched per campaign per sweep. Eleven emails for eleven conversations is the fastest way to get filtered to a folder.
• Count and best score in the subject, so it's triageable from a phone notification.
• Dedupe is a column on the row, not a time window — comparing found_at against a last-alerted timestamp double-sends anything that arrives mid-compose.
• Marked alerted only after the send succeeds. The other order loses a batch permanently to one SMTP hiccup.

R2 — keyword matching drops the best requests

"Our checkout keeps timing out whenever we run a promo. Anyone recommend a fix?"

Exactly the buyer a load-testing campaign wants. Contains none of its keywords — because people with a problem describe the problem, not the product category that solves it.

A campaign can now describe what it sells in prose, and a keyword miss becomes a question rather than a verdict. The path is deliberately narrow:

• only judges signals the cheap path already scored above the bar
• capped per sweep, so an index returning 300 results can't become 300 model calls
• cannot overturn a stated no or rescue somebody advertising their own services
• must quote the words that decided it — a keyword match is wrong in obvious ways; a model judging relevance is wrong in ways nobody sees unless it says why

Deliberately not copied

• Facebook/Instagram monitoring. Neither carries meaningful public buying intent, neither is readable without access we don't have. Listing them would be coverage we can't deliver.
• Per-keyword subscription tiers. A second quota beside credits makes the cost of a run unpredictable — the one thing this week's billing work was for.

Checks

• tsc --noEmit clean · 1,100 tests pass (13 new) · build compiles
• Migration applied to prod

Remaining in the PRD: R3 outcome tracking, R4 per-source analytics, R5 reply prompts, R6 Slack/Telegram, R7 keyword buckets.

🤖 Generated with Claude Code

[github-actions]

vu1nz Security Review

0 finding(s) in PR #?

No security issues found.