List the catalogue the directory actually holds

The README described three seed templates and said the set stays small enough to hold in your head.
It holds twenty-seven across nine kinds of work, and its table named three of them — so the file a
reviewer reads before adding one was describing a directory that no longer exists.

The table now carries every file with the category it files itself under, which is also the quickest
way to see that the closed vocabulary is populated rather than aspirational.

Author: jerelvelarde

examples/templates/README.md

@@ -5,20 +5,49 @@ depends on, the capabilities it *asks* for, and a ceiling on what it may do. It
 endpoint, no credential and no grant. See [docs/bot-templates.md](../../docs/bot-templates.md) for
 the format.
 
-These three are the seed of the catalogue. The `Dockerfile` copies `examples/` into the image, so
-they travel with a deployment that has no network at all and can be imported by opening one and
-pasting it in. They are also the worked examples: a template somebody writes by hand is read against
-these.
+There are 27 of them, across the 9 kinds of work a category may name. The `Dockerfile`
+copies `examples/` into the image, so they travel with a deployment that has no network at all and
+can be imported by opening one and pasting it in. They are also the worked examples: a template
+somebody writes by hand is read against these.
+
+The set is deliberately varied rather than uniform — some ask for no connectors at all, one runs
+remote, one buys exactly a single write — so that it demonstrates the boundary vocabulary instead of
+repeating one shape 27 times. What each one is really carrying is a rule about judgement: what it
+must not conclude, and what it says when the evidence is thin.
 
 They are deliberately **written rather than exported**. A template may not carry `type: built_in`
 with a `system_prompt`, so exporting one of the shipped fintech Bots is not a faithful round trip —
 which means the first catalogue entries have to be written, and that is what these are.
 
-| File | What it is for |
-| --- | --- |
-| `research-desk.openbot.yaml` | Reads around a question and writes a brief that names every source. |
-| `ticket-triage.openbot.yaml` | Decides what an incoming ticket is and drafts a reply for a person to send. |
-| `competitor-watch.openbot.yaml` | Watches a fixed list of pages and says what actually changed. |
+| File | Category | What it is for |
+| --- | --- | --- |
+| `account-health.openbot.yaml` | customer-success | Reads one account's own record and says what the evidence supports about its health, naming what is missing rather than filling it in. Gives no score. |
+| `account-research.openbot.yaml` | sales | Prepares one company before a meeting from what they publish and what we already recorded, and keeps the two apart so nobody repeats a claim back as a fact. |
+| `beta-adoption.openbot.yaml` | product | Says which accounts have actually used a feature in the beta, against the roster of who was let in, and keeps "did not use" apart from "not measured". |
+| `campaign-brief.openbot.yaml` | marketing | Turns a request for a campaign into a brief somebody can build from, and marks what nobody has answered as an open question rather than filling it in. |
+| `chief-of-staff.openbot.yaml` | general | Keeps a written record of what people said they would do, and says what has slipped without chasing anybody. |
+| `claims-desk.openbot.yaml` | marketing | Checks every factual claim in a marketing draft against a document we can name, and refuses to approve, rewrite or treat our own earlier copy as evidence. |
+| `competitor-watch.openbot.yaml` | marketing | Watches a fixed list of competitor pages, says what actually changed since last week, and refuses to dress a change up as a conclusion. |
+| `escalation-watch.openbot.yaml` | customer-success | Watches open threads for a commitment that has been missed, and drafts the handover for a person to send. Escalates nothing itself. |
+| `expense-review.openbot.yaml` | operations-finance | Reads one expense claim against the policy as it is written, quotes the clause that applies, and leaves the decision to a person. |
+| `feedback-digest.openbot.yaml` | product | Groups what customers actually said into themes, keeps every theme attached to the sentences that produced it, and never promotes a suggestion into a commitment. |
+| `inbox-first-pass.openbot.yaml` | life | Reads a morning's messages where they already live and says which few need you today. Answers none of them. |
+| `interview-notes.openbot.yaml` | recruiting | Turns what an interviewer says into a written record: the question, the answer and what was observed, kept apart from what was concluded. Asks for nothing. |
+| `invoice-chaser.openbot.yaml` | operations-finance | Works out which invoices are actually overdue, drafts the follow-up a person sends, and can hold the weekly chase to a schedule. |
+| `meeting-follow-ups.openbot.yaml` | general | Turns one set of meeting notes into the follow-ups actually in them, each with the line it came from, and leaves the unowned ones unowned. |
+| `onboarding-buddy.openbot.yaml` | recruiting | Answers a new starter's questions from the written handbook, quotes the passage, and books the first fortnight's check-ins. Hands anything with a legal edge to a person. |
+| `oncall-handover.openbot.yaml` | engineering | Writes the handover one on-call shift leaves the next: what fired, what is still open, and what the log does not say. |
+| `outbound-desk.openbot.yaml` | sales | Reads what we already know about an account and drafts a first-touch note for a person to send. Never sends anything, and never invents a reason to write. |
+| `paid-media-desk.openbot.yaml` | marketing | Reports what paid spend did over a stated window, from an agent your own team runs, and refuses to explain a movement or recommend a budget. |
+| `product-performance.openbot.yaml` | product | Answers questions about how the product is doing through your own metrics service, and names the definition and the window behind every number it gives. |
+| `release-notes.openbot.yaml` | engineering | Turns a list of merged changes into release notes a customer can read, and counts the ones with no visible effect rather than dressing them up. |
+| `renewal-prep.openbot.yaml` | customer-success | Assembles the pack for a renewal conversation and refuses to write the ask. Runs on a service you host yourself, so the whole conversation leaves this deployment. |
+| `renewal-watch.openbot.yaml` | sales | Reads renewal dates and notice periods off the contract, lists what is coming, and refuses to turn a quiet account into a risk score. |
+| `repro-desk.openbot.yaml` | engineering | Follows a bug report step by step on a site you name, says whether it reproduces, and does not diagnose it. |
+| `research-desk.openbot.yaml` | general | Reads around a question and writes a short brief that names every source it used. |
+| `talent-scout.openbot.yaml` | recruiting | Reads a role's written requirements, then finds published evidence against each one and gives the address it was read on. Does not rank people and contacts nobody. |
+| `ticket-triage.openbot.yaml` | customer-success | Reads an incoming ticket, works out what it actually is, and hands a person a draft reply and a reason. Never answers the customer itself. |
+| `vendor-review.openbot.yaml` | operations-finance | Assembles what is actually known before a renewal — the terms we signed, what we spent, and what the vendor publishes today — and recommends nothing. |
 
 ## Review rules