Rewrite a pending task's message

Who approved it, as a name. Null when nobody has, or the person could not be resolved.

Why a message that could have sent unattended was put in front of a person.

One vocabulary for every gate between an agent writing a message and the message leaving: what the agent itself said about the draft (LOW_CONFIDENCE, FINDINGS, INSUFFICIENT_DATA, POLICY), what the sender or the cadence asked for (ALWAYS_ASK, SEQUENCE_MANUAL, STEP_REQUIRES_APPROVAL), what the pacing rules refused (DAILY_CAP, CONSECUTIVE_CAP, THREAD_CAP, HOURLY_CAP, ORG_DAILY_CAP), what the mailbox or the thread made unsafe (UNVERIFIED_SENDER, DIFFERENT_SENDER, LABEL, QUESTION, STALE_INBOUND, SEQUENCE_GONE), and what the counterparty turned out to be (BOT_QUESTION, AUTOMATED_COUNTERPARTY, AUTHENTICATION, LANGUAGE, LINKEDIN_ONLY). KILL_SWITCH is the platform switch that stops unattended sending outright, and SEND_INTERRUPTED an unattended reply whose send never got as far as a provider. SENDER_SWITCHED_TO_MANUAL is a LinkedIn step queued unasked that went back to a person because its sender switched to manual before it ran.

Stored on the task so the queue can say why each row is there, and so "why did nothing send" is answerable by counting rows rather than by reading logs.

The rendered message as it will send: an email body, a note, a LinkedIn message.

Which channel performs the task once it is allowed to run.

How sure the agent was that its message could go out unedited: high (a person could send it unread), medium (worth a glance) or low (needs work). Null when no agent wrote the message.

The number behind confidence: the agent's own score for its message, from 0 to 100. Read from the agent's message as it stands now. Null when no agent wrote the message, or the agent gave no score.

The contact's company, from whichever column of the audience table holds it ('Company', 'Company Name', 'Organization' and the like). Null when the table has none or leaves it blank.

The address, profile slug or number behind the label, so a row can show the name and still say who that is. Equal to the label where no name was found; null for housekeeping.

The contact as the queue names them: a name, else an address or a profile URL.

The contact's job title, from whichever column of the audience table holds it ('Job Title', 'Title', 'Position' and the like). Null when the table has none or leaves it blank.

When the channel would run it if nobody were in the way.

Whether the queue may rewrite the message before approving: pending, and a kind that carries text.

Whether a check found a problem in the agent's message (any finding at warning strength), whatever ask_reason says: a task held for a setting can still carry a message with a placeholder that never filled in. False when no agent wrote the message.

The LinkedIn action kind slug or email.

What the person wrote when settling it: why they rejected it.

The message the Outbound Agent wrote for this task. Null when a person wrote the words, and on a LinkedIn step that carries no message.

How it ended: the channel's own terminal word.

A separate AI review's score for the agent's message, from 0 to 100: how likely a busy person in the recipient's role is to read it to the end and reply. Null when no agent wrote the message, or when it was not reviewed.

The agent's own note on its message: what it built the message on and what it was unsure about. Null when no agent wrote the message.

For an email that replies in the contact's existing thread, the subject it will go out with, such as 'Re: Quick question'. Null for a first email, which carries its own subject, and for anything that is not an email.

For a LinkedIn task still waiting on a person: whether a browser is checked in for its sender right now, so approving it would run it soon. False means an approved step waits until a browser is open. Null for every other task.

Null when nobody in particular sends it, or the sender was deleted.

Null for housekeeping work, or for a sequence since deleted.

Its sequence is paused right now: approving the task records the approval, and the step waits for Resume sequence before it runs, on either channel.

Where one task stands.

PENDING is waiting on a person. SCHEDULED is approved, or never needed approval, and is waiting on the channel to execute it: the browser to run the LinkedIn step, the engine to send the email. The task is where a LinkedIn step's approval lives: while it is PENDING no step exists, and the step is queued only once it is SCHEDULED. A LinkedIn task nobody is asked about (a sender on automatic, a housekeeping read, a Try click) is BORN scheduled, pointing at the step it queued. Who approved it is a stamp on the row, not a status of its own.

DONE, REJECTED and CANCELLED are terminal. DONE is the channel executed it; REJECTED is a person refused it; CANCELLED is the channel gave up on it (failed, skipped, expired), or the contact was stopped underneath it.

The step as the builder names it; null when the step no longer exists.

The step's customer-visible ordinal, zero-based; null with the name.