connect-whatsapp-capture6.1 KB
View saved version →
---
name: connect-whatsapp-capture
description: Connect the user's own WhatsApp number to Werify, measure available chat history, or configure capture of selected chats. Use for pairing, previewing history, and starting or stopping new-message or earlier-history capture.
---
# Connect WhatsApp and choose what to capture
Pairing, measuring history, and capturing messages are separate choices. Use the connected
Werify MCP tools only for the steps the user requested. Explicit user instructions take
precedence over workflow guidance; they do not bypass server authorization or confirmation.
Opening this onboarding skill alone does not authorize pairing or capture. Establish the
user's goal if it is missing; an existing archive can be read without connecting a number.
## Connect and choose chats
Check `list_connected_numbers` before pairing. For a new connection, obtain the user's number
with country code and call `connect_whatsapp_number` if available. Relay its pairing code and
phone instructions, then check the number's status with `list_connected_numbers`. A code is
not proof of connection: wait for `ready` before selecting chats. Stop on expiration or an
actionable error and follow the returned next step. Do not repeatedly restart pairing.
If the number belongs to another Werify account, relay the refusal: the user moves the number to
this account in the web app, through the link it returns. You cannot move it.
Use `list_capturable_chats` with the returned phone, name/activity filters and pagination.
Resolve duplicate names using returned handles. `chat_id` identifies a phone chat;
`conversation_id` identifies an archive for read tools. Do not substitute one for the other.
Select only the chats the user requested. Activity and direct-chat filters do not establish
that a contact is a customer, a lead, or a newly created conversation.
## Measure without capture
Use `start_history_preview` before or after capture, with the phone and chat handle from
`list_capturable_chats`, following its current schema. If it is unavailable or a stale tool
definition is rejected, use the returned web link or https://werify.app/conversations for that
operation. Do not try retired preview names or repeat an unsupported call.
Measuring reads metadata and can cause phone sync notifications; it does not save messages or
media into the archive. Relay `next_step`, including guidance to keep the phone online with
WhatsApp open. At `queued` or `measuring`, report progress, not completion. Check again
after 1–3 minutes if the host can continue the task; otherwise explain when the user can check.
Do not poll in a tight loop or promise a background follow-up the host cannot perform.
With `start_history_preview`, resume a stopped, failed or unanswered measure with
`continue: true` only at the user's request.
Use ready `available_from` for the earliest measured day. `reached_back_to`, `messages_seen`
and null dates do not prove a finished measurement or a new chat. For a permitted batch,
inspect every entry and individual error. A request to measure never authorizes capture.
## Capture only the chosen scope
- **New messages:** use `start_live_capture` for the agreed chats and transcription choice.
It starts from the conversation's local midnight, including earlier messages today.
If the user requires strictly after this instant, explain the available cutoff and wait for
agreement to include today's earlier messages before proceeding.
When only `start_capture` is available, use its forward-only path without history fields;
do not promise identical cutoff semantics. No history preview is needed for this request.
- **Earlier history:** show the selected preview option's dates, counts, transcription choice
and coverage, and obtain agreement. Use `start_history_capture` with the exact served date;
for the `all_history` option whose `from_date` is null, pass the string `"all"`, not null.
Enable `with_live_capture` only if new-message captu
Preview of the first 4,000 characters. Read full instructions →
Referenced files: 1 — file references are metadata, not package contents.
read-whatsapp-archive5.14 KB
View saved version →
---
name: read-whatsapp-archive
description: Search, read, and summarize WhatsApp conversations in the user's connected Werify archive, including available voice-note transcripts and attachments. Use for questions about archived messages, decisions, dates, or shared files.
---
# Read a WhatsApp archive
Use the connected Werify MCP tools within the user's requested scope. Explicit user instructions
take precedence over this workflow guidance. Reading an archive does not authorize connecting a
number, starting capture, obtaining earlier history, or sending a message.
## Find and read
1. Resolve the conversation with `search_conversations`. Its query matches conversation names,
not message content. Use returned identifiers; disambiguate duplicate names with the user.
To find a person, their number or every conversation they appear in, use `list_contacts` if
it is in your tool list: it matches a name or a number and returns each conversation's id.
2. For a period summary or all decisions, use `get_conversation` with the requested local
calendar dates. Follow `next_cursor` until null within that window. Pages arrive newest-first;
messages inside each page are chronological. Reconcile later decisions with earlier ones.
3. For a specific phrase or fact, use `search_messages`. When the user does not say which
conversation it is in and the schema makes `conversation_id` optional, leave it out: the search
spans every conversation, and each match names its conversation. Otherwise search within the
resolved conversation. It matches literal substrings, including available audio transcripts,
not semantic similarity, so keep the query to the few words that must appear. On every page
send the same `query`, and the same `conversation_id` when you passed one, plus the cursor.
A keyword search alone cannot establish every decision in a period.
Prefer the callable tool's current schema over remembered arguments. Tools can be reviewed
independently, so only call names actually present in your tool list.
## Voice notes and files
Read voice-note content from available inline transcripts in `get_conversation`. A missing
transcript is a coverage gap; do not infer what the speaker said.
For requested files, use `list_attachments`, filter by the requested type and dates, and page
until `next_cursor` is null, sending the same `conversation_id` with each cursor when you passed
one. If its schema has them, pass the words the user gives for the file as
`query`, which matches the file name and the caption, and leave out `conversation_id` when the
user does not say which conversation the file is in: the list then spans every conversation.
When the user has more than one connected number and says which one, pass it as `phone` to
`search_messages`, `list_attachments` and `list_contacts` if their schemas have it. For PDFs, use
`type: "document"` and identify PDFs from returned file metadata; `pdf` is not a supported type.
It lists attachments whose bytes Werify holds: absence does not prove a file was never sent.
Listing files does not require downloading them.
When the user asks to inspect or download selected files, or explicitly requests temporary
links, use `get_attachment_file` only for their returned `media_id` values. If its current schema
accepts lists, group requested IDs into batches of at most 100, removing duplicates while
preserving order. A list, even with one ID, returns `results`; inspect every entry and report
per-file errors without discarding successful entries or claiming all files succeeded. A string
returns one download. If the available schema accepts only a string, use individual calls.
Listing alone must not issue temporary download links. Fetch bytes only when inspection or
download was requested and the host supports it; asking for links alone does not request a fetch.
These URLs are bearer links valid until `expires_at`, not host file identifiers. Do not publish
them in shared artifacts or as
Preview of the first 4,000 characters. Read full instructions →
Referenced files: 1 — file references are metadata, not package contents.