Connection Pipeline Tracker

Connection Pipeline Tracker

Once a brand has outreaches in flight, the question stops being "who do I reach?" and becomes "where does everything stand right now, and what needs me today?" This skill answers that: it reads the brand's live outreaches and lays them out as a pipeline board grouped by lifecycle stage, with the action-needed buckets (interested, question, stalled) pulled to the top. It is strictly read-only — it calls only the two free connection-read tools and never reaches anyone, never charges, and never shows or infers contact info. For any brand operator running an active outreach program who wants a status-at-a-glance view.

Read first: ${CLAUDE_PLUGIN_ROOT}/shared/conventions.md (tool schemas, credit prices, the nine conventions) and ${CLAUDE_PLUGIN_ROOT}/shared/connection-flow.md (THE connection contract — tool schemas, entitlement/graceful-degradation, and the privacy invariants this board lives inside).

This is the board, not the action. It does not draft replies, does not reach creators, does not price. If a reply needs answering, point to reply-triage; if a stalled outreach needs a follow-up nudge, that's the follow-up flow / cast-and-connect; to reach NEW creators, use the connection-enabled casting skills. This skill only reads list_connections (free) and get_connection_status (free) and renders the state.

Inputs to collect

• Status filter (optional) — if the user only wants one stage ("just show me who replied", "what's still pending"), pass it through as the list_connections status filter. Default: pull everything (no filter) so the full board renders.
• Campaign focus (optional) — if the user names a campaign type ("just my spring-launch outreaches"), filter the rendered board to that campaign_type client-side (the list tool filters on status, not campaign — so fetch all, then narrow in the board).
• Detail depth (optional) — by default, call get_connection_status only on the outreaches worth detail (everything in an action-needed bucket, plus anything mid-sequence where the next touch matters). The user can ask for "full detail" to expand get_connection_status to every row. Both are free, so this is about output density, not cost.

Never re-ask anything the list already tells you. No credit mode applies — the whole skill is free.

Flow

Step 0 — Entitlement / empty-state pre-check (graceful, never an error). The brand's plan may not include creator_connections, or may simply have no outreaches yet. Either way this skill degrades gracefully and never errors:

• If list_connections comes back as the refused/non-entitled envelope (per connection-flow.md), say plainly: "Connections aren't on your plan, so there's no outreach pipeline to show — here's how to upgrade." Stop there; do not fabricate a board.
• If the plan is entitled but list_connections returns empty, say: "No outreaches found — you haven't reached any creators yet. When you do (via a connection-enabled casting skill), they'll show up here as a board."

Step 1 — Pull the outreaches (free). One call, newest-first:

list_connections { "limit": 100 }

If the user asked for a single stage, scope the call:

list_connections { "status": "replied_question", "limit": 100 }

Each row returns conn_ref, status, campaign_type, and next scheduled touch (no contact info — none is returned, by design). If a campaign focus was given, drop rows whose campaign_type doesn't match before rendering.

Step 2 — Detail the ones worth detailing (free). For every outreach in an action-needed bucket (interested / question / stalled), and any mid-sequence row where touch progress matters, call:

get_connection_status { "conn_ref": "<ref from step 1>" }

→ lifecycle status, touches sent (of 3), latest reply classification if any, next scheduled touch. Still no contact details, still free. (In "full detail" mode, run this for every row.) These calls are free, so detail liberally — but don't call it on rows where the list already told you everything (e.g. a fresh queued row with nothing to expand).

Step 3 — Bucket by lifecycle stage and compute age. Group every outreach into its stage and, within each, sort by what the operator should act on first. The lifecycle order:

queued → sending → delivered →
  replied_interested · replied_question · replied_declined →
  accepted_inapp →
  opted_out · expired

For each outreach capture: conn_ref, campaign type, touches sent (of 3), next scheduled touch, age (time since the outreach was created / first sent — derive from the timestamps the tools return; if a precise created timestamp isn't returned, say "age unknown" rather than guessing), and the engagement block (opened_at / clicked_at / click_count / tracking_enabled — see connection-flow.md for the exact shape and the framing rules).

Step 4 — Lift the action-needed buckets to the top. Four buckets are where the operator's attention belongs, so they lead the board:

• Interested — awaiting your move (replied_interested): the creator said yes-ish; the brand needs to advance it. The double-opt-in intro is the matchmaker's to make on a confirmed yes — frame these as "ready for you to take forward," not "here's their email."
• Question — needs a reply (replied_question): the creator asked something; a reply is owed. → point to reply-triage to draft it.
• Engaged, not replied (delivered, no reply classification, but the engagement block shows a click — or, weaker, an open): the warm-lead surface. A click is reliable intent (lead with clickers); an open alone is directional only (MPP inflates opens — say so, don't oversell). Rows with tracking_enabled: false never appear here as "no engagement" — they're unmeasured, not cold.
• Stalled — delivered, no reply, Day-7 approaching (delivered with no reply classification and the next/last touch nearing the Day-7 end of the 3-touch sequence): the sequence is about to lapse. → point to the follow-up flow if the brand wants one more nudge before it expires. (An outreach can sit in both engaged-not-replied and stalled; show it in engaged-not-replied — the warmer read — and note the sequence is ending.)

Everything else (queued, sending, mid-sequence delivered, accepted, declined, opted-out, expired) renders below as the steady-state pipeline.

Deliverable

A clean, scannable markdown board (mirrors talent-scout's what's-changed style — lead with what needs action, steady state below):

# Outreach Pipeline — <brand / engagement>
_Creatorland Data · live board as of <date> · <campaign focus, if any>_

## Needs you now (<A>)

### Interested — awaiting your move (<i>)
| conn_ref | Campaign | Touches | Replied | Age |
|---|---|---|---|---|
| <ref> | <campaign_type> | <n>/3 | interested | <age> |
> Ready to take forward — Creatorland makes the double-opt-in intro on a confirmed yes.

### Question — needs a reply (<q>)
| conn_ref | Campaign | Touches | Asked at | Age |
|---|---|---|---|---|
| <ref> | <campaign_type> | <n>/3 | <when> | <age> |
> → Draft replies with **reply-triage**.

### Engaged, not replied (<e>) — warm leads
| conn_ref | Campaign | Touches | Clicked | Opened | Age |
|---|---|---|---|---|---|
| <ref> | <campaign_type> | <n>/3 | <clicked_at or —> | <opened_at or —> | <age> |
> Clicks = reliable intent; opens are directional only (privacy prefetches, e.g.
> Apple Mail, inflate them). Tracking since 2026-08-18 — earlier sends are
> unmeasured, not cold.

### Stalled — delivered, no reply, Day-7 approaching (<s>)
| conn_ref | Campaign | Touches | Next touch | Age |
|---|---|---|---|---|
| <ref> | <campaign_type> | <n>/3 | <date or "sequence ending"> | <age> |
> → One more nudge before it expires? Use the follow-up flow.

## Pipeline (steady state)

| Stage | Count | conn_refs |
|---|---|---|
| queued | <n> | <refs> |
| sending | <n> | <refs> |
| delivered (mid-sequence) | <n> | <refs> |
| accepted (in-app) | <n> | <refs> |
| declined | <n> | <refs> |
| opted out | <n> | <refs> |
| expired | <n> | <refs> |

## At a glance
- **<T> outreaches total** · <A> need action · <live> live · <closed> closed.
- Reply rate so far: <r>/<delivered> delivered have replied (a count, not a promise).

---
Status is brand-safe; contact is not — this board shows where each outreach
stands, never a creator's contact details (none are ever returned by these
tools). On a confirmed yes, Creatorland makes the double-opt-in intro.
Provenance: Creatorland Data MCP · list_connections + get_connection_status
(read-only) · <date>.
Credits used this run: 0 (both connection-read tools are free).

> Live-artifact tip: in a harness with live artifacts (e.g. Cowork), this board
> is a strong auto-refresh candidate — pin it as an artifact and re-run the read
> to keep the pipeline current without re-asking. (Honest constraint: there's no
> server-side push/change-feed; "refresh" means re-reading the free tools on
> demand or on a schedule, not a live socket.)

If the entitlement or empty-state branch fired, the deliverable is just that one honest line (plus the upgrade pointer for the non-entitled case) — never a fake board.

Honesty rules

• Read-only, always. This skill calls only list_connections and get_connection_status. It never calls request_creator_connection, never reaches a creator, never advances a sequence — it renders state. If the user wants to act, it points to the skill that acts.
• Status is brand-safe; contact is not. A connection's existence + lifecycle status is safe to show; a creator's contact details are never returned by these tools and never appear or are inferred in this board (convention 7 + connection-flow privacy invariants). The "interested" affordance is "ready to take forward via Creatorland's double-opt-in intro," never an address.
• Never promise a reply or a yes. Counts and stages are facts; outcomes are the creator's. "Reply rate so far" is a tally of what's happened, not a forecast. "Stalled" describes the sequence state, not a verdict on the creator.
• Engagement framing (connection-flow.md rules). Clicks are the headline intent signal; opens are directional only (automated privacy prefetches, e.g. Apple Mail, inflate them). tracking_enabled: false means "sent before tracking was enabled (2026-08-18)" — present it as unmeasured, never as "0 opens / 0 clicks".
• Graceful degradation, never an error. No creator_connections entitlement → "not on your plan, here's how to upgrade." No outreaches → "no outreaches found yet." Empty stage filter → "nothing in that stage right now." Always a clean, honest line — never a thrown error or an invented row.
• Free means free. Both read tools cost 0 credits; say so. Don't imply this skill spends anything, and don't let "full detail" mode read as a paid upsell — get_connection_status is free at any volume.

Credit footprint

thorough: 0 credits · thrifty: 0 credits — list_connections and get_connection_status are both free, so this skill is free at any size and in any mode. It reaches no one and runs no market-intel or profile calls.