Suppression Aware Prescreen

Suppression-Aware Pre-Screen

Before a brand commits a list to a campaign, two kinds of names should never make it to the matchmaker: creators who already opted out of this brand's outreach, and creators already in an ACTIVE sequence for this brand (a repeat would be refused, uncharged, and is a harassment / sender-reputation risk). This skill takes the list you're about to reach, reads the brand's connection records, and partitions it into REACHABLE / SKIP — OPTED OUT / SKIP — ALREADY ACTIVE — telling you who and why for every skip, and handing back a clean set ready for cast-and-connect or conflict-safe-connect. It is strictly read-only: it calls only the free list_connections, never request_creator_connection, never reaches anyone, never charges. For any brand operator about to commit to a reach.

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 — the suppression / active-sequence pre-screen, graceful degradation, and the privacy invariants this screen lives inside).

This is the standalone version of the suppression / active-sequence pre-flight that the reach skills embed. Run it before you commit a list to a campaign, so you size the real reach (and its cost) on the clean set. Siblings: connection-pipeline-tracker is the FULL board of everything in flight (not a list-screen); reply-triage handles replies you've received. This skill takes a candidate list IN and returns the reachable subset OUT — it does not reach, does not draft, does not price the reach. Hand its clean set to cast-and-connect / conflict-safe-connect to actually reach.

Inputs to collect

• The list to screen (required) — the creators you're about to reach, in any format: handles, an uploaded sheet, or an in-session shortlist (e.g. the output of brief-to-shortlist). Map each to one identifier so it can be matched against the connection records. If no list is given there is nothing to screen; ask once for it.
• Brand / campaign scope (optional) — the connection records are already scoped to the brand's account, so no brand id is needed. If the user names a campaign ("just for the spring launch"), the active-sequence check is narrowed to outreaches whose campaign_type matches; otherwise an active sequence for ANY campaign of this brand counts as already-active for the reach-refusal purpose.

Never re-ask anything the list or the connection records already tell you. No credit mode applies — the whole skill is free (list_connections 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 nothing to pre-screen against yet — every name on your list is reachable as far as our records go. Here's how to upgrade to start reaching." Treat the whole list as REACHABLE (there is no suppression ledger or active sequence to subtract) and stop — do not fabricate skips.
• If the plan is entitled but list_connections returns empty, say: "No prior outreaches on record — nothing to pre-screen against, so the full list is clear to reach." Again, treat the whole list as REACHABLE.

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

list_connections { "limit": 100 }

Each row returns conn_ref, status, campaign_type, and the next scheduled touch — no contact info (none is ever returned, by design). If more than 100 outreaches exist, page by the lifecycle statuses that matter to this screen so nothing is missed:

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

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

(repeat for the other ACTIVE statuses below). If a campaign scope was given, keep only rows whose campaign_type matches for the active-sequence test; opt-outs count brand-wide regardless of campaign.

Step 2 — Classify the connection records (no tool call). From the rows:

• OPTED-OUT set = every outreach with status opted_out. These creators asked this brand to stop — the suppression ledger. Brand-wide.
• ACTIVE set = every outreach in a live, mid-sequence state where a repeat request would be refused and not charged: queued, sending, delivered, replied_question, and replied_interested (the sequence / conversation is still open). Treat accepted_inapp as active too — they're already engaged; re-reaching is noise. Not active for this purpose: replied_declined and expired (the sequence is closed — re-reaching is a judgement call, not an automatic refusal; surface them as a note, not a skip).

Step 3 — Partition the input list against those sets (no tool call). For each creator on the list to screen, match against the OPTED-OUT and ACTIVE sets:

• In OPTED-OUT → SKIP — OPTED OUT (cite the conn_ref).
• Else in ACTIVE → SKIP — ALREADY ACTIVE (cite the conn_ref, status, and campaign — "a repeat would be refused and uncharged").
• Else → REACHABLE (clear of suppression and active sequences in our records).

Compute the credit savings: each avoided reach = 10 credits not wasted (request_creator_connection is 10 credits and an already-active repeat would be refused-but-still-a-mistake; an opt-out reach is a reputation cost). Savings = 10 × (count of SKIP rows).

Deliverable

A pre-screen partition in markdown:

# Outreach Pre-Screen — <list / campaign name>
_Creatorland Data · screened against this brand's connection records · <date>
<· campaign scope, if any>_

## Partition
| Creator | Verdict | Why | conn_ref |
|---|---|---|---|
| @<handle> | ✅ REACHABLE | clear in our records | — |
| @<handle> | 🚫 SKIP — OPTED OUT | this creator opted out of your outreach | <conn_ref> |
| @<handle> | ⏳ SKIP — ALREADY ACTIVE | already in a live <campaign_type> sequence (<status>) — a repeat would be refused & uncharged | <conn_ref> |
<one row per creator on the list>

## Reachable set (clear to hand off)
@<handle>, @<handle>, … &nbsp;(<R> of <T>)
> Ready for **cast-and-connect** or **conflict-safe-connect** to actually reach.
> (This skill does not reach — it returns the clean list.)

## Skipped (and why)
- 🚫 **Opted out (<o>):** @<handle> — asked your brand to stop; do not re-reach.
- ⏳ **Already active (<a>):** @<handle> — live <campaign_type> sequence, <status>; track it in `connection-pipeline-tracker`, don't re-reach.
<If none: "Nothing skipped — the whole list is clear in our records.">

## Worth a look (not skipped)
<Any list creators matching a CLOSED outreach — replied_declined / expired —
listed as a judgement call, not an automatic skip; or "None.">

## Credit savings
Pre-screening held back <S> reach(es) you'd otherwise have committed —
**~<10×S> credits (~$<0.25×S>) not wasted**, plus the sender-reputation cost of
re-contacting an opted-out or already-active creator avoided.

---
A clean screen means **none of these are suppressed or in an active sequence in
our records** — not an absolute guarantee. This is a read-only pre-flight; it
reaches no one. Contact details are never returned by these tools and never
appear here.
Provenance: Creatorland Data MCP · list_connections (read-only, free) · <date>.
Credits used this run: 0 (`list_connections` is free; nothing was reached).

If the entitlement or empty-state branch fired, the deliverable is the honest one-liner plus the full list marked REACHABLE — never a fabricated skip.

Honesty rules

• A clean screen is "none are suppressed / active IN OUR RECORDS" — never an absolute guarantee. The connection records are this brand's outreach history, not every interaction a creator has ever had; phrase it that way everywhere.
• Read-only, always. This skill calls only list_connections. It never calls request_creator_connection, never reaches a creator, never advances a sequence, never charges. If the user wants to act on the clean set, it points to the skill that reaches — it does not reach itself.
• A skip is a protection, not a verdict on the creator. "Opted out" and "already active" describe the record, not the creator's worth — frame them as "don't re-reach right now," and route already-active ones to the pipeline tracker rather than dropping them silently.
• Closed ≠ active. replied_declined and expired are surfaced as a judgement call ("worth a look"), never auto-skipped — re-reaching a lapsed sequence is the operator's decision.
• Graceful degradation, never an error. No creator_connections entitlement or no outreaches on record → "nothing to pre-screen, the list is clear," with the whole list marked REACHABLE — never a thrown error, never an invented skip.
• Credit savings are avoided waste, stated honestly. Each skip saves the ~10 credits an avoided reach would cost (an already-active repeat is refused anyway; an opt-out reach is a reputation cost) — a real saving, not a discount.
• PII invariant (convention 7 + connection-flow privacy). The screen returns counts, conn_refs, and status only — no contact info, regardless of source. A connection's existence + status is brand-safe; a creator's address is not, and never appears or is inferred here.

Credit footprint

thorough: 0 credits · thrifty: 0 credits — list_connections is free, so this skill is free at any list size and in any mode. It reaches no one, runs no profile or market-intel calls, and exists precisely to save the 10×S credits a list of S already-suppressed / already-active names would otherwise waste.