<summary>Query the posts an agent's LinkedIn monitors discovered — this agent's own
monitor-scoped feed, not the account-global post analytics. Two modes:
mode="search" (default): individual post rows.
mode="stats": per-status counts.
Each row is one post a monitor surfaced, with the comment Sliq drafted for it
(`draft_comment`) and the post's lifecycle `status`: 'discovered' (claimed,
pre-draft), 'filtered' (the monitor's post filter set it aside, so no action ran
on it), 'queued' (drafted + enqueued), 'commented' (comment sent),
'rejected' (user declined the draft), 'skipped' / 'skipped_no_credit' (no
comment drafted), 'captured' (a capture ran on it — its engagers, or its
author when a person — and it wasn't commented), 'surfaced' (no comment and no
capture ran — a surface-only monitor, or a company page's or the user's own post
on an author-only monitor — just surfaced in the feed to act on manually).
`author_captured` is whether author capture ran on the post's author, a person;
the monitor's list holds them unless the user removed them.
`filter_reason` is the post filter's one-sentence reason for keeping or setting
aside the post ('' when the monitor has no filter). Read it beside the filtered
and kept posts when the user wants to tune their monitor's post filter.
`data` holds the author's `headline` and `public_identifier`. Link a post's author
with the derived `author_profile_url` on each search row — a person's profile, or
the company page when a company posted. It is computed after the query runs, so it
can't appear in `where_clause` or `order_by`.
`reaction_status` is the independent reaction track: '' (no reaction),
'queued' (a reaction enqueued), 'reacted' (sent), 'rejected' (user declined it),
'skipped' (deduped) — disjoint from `status` since one post can be both
commented and reacted to. `reaction_type` is which reaction landed
(like/celebrate/support/love/insightful/funny, '' until sent).
Columns: id, monitor_id, post_urn, post_url, author_name, author_provider_id,
data, post_text, reaction_count, comment_count, posted_at, status, filter_reason,
draft_comment, reaction_status, reaction_type, author_captured, created_at. Pass a row's `id` to
generate_monitored_post_comment to draft + queue a Sliq comment for it.
Pass `agent_id` to scope to that agent's monitors — omitting it reads every
monitored post across all the user's agents. To act on recent posts (e.g. queue
a reaction on this agent's fresh finds), pass `agent_id` and filter on
`posted_at`; this keeps actions on the agent's own scoped feed rather than the
account-wide analytics.
This is the monitor-scoped feed. `query_linkedin_posts` is a different table —
the account-global analytics of the user's own posts plus any post they've
fetched engagers for, with no agent/monitor dimension. Use that for "how did my
posts perform"; use this for "what did this agent's monitors find".</summary>
<returns>
<description>In search mode, {count, truncated, items array}.
In stats mode, {total, by_status: {discovered: N, commented: M, ...}}.</description>
</returns>