Competitor Watch

Competitor Creator Watch

Answers "who is creating around the brands we compete with?" — in two layers: (1) creators the corpus links to the named competitor brands via brand affinities/affiliations, and (2) whitespace: similar creators the corpus does NOT link to those competitors, i.e. casting targets a rival hasn't visibly claimed. For brand-side managers sizing up a rival's creator strategy and agency strategists pitching whitespace slates.

Read first: ${CLAUDE_PLUGIN_ROOT}/shared/conventions.md (tool schemas, credit prices, the nine conventions). This skill honors thrifty/thorough credit modes (${CLAUDE_PLUGIN_ROOT}/shared/credit-modes.md), Refusal Recovery (${CLAUDE_PLUGIN_ROOT}/shared/refusal-recovery.md), and the Freshness Gate (${CLAUDE_PLUGIN_ROOT}/shared/freshness-gate.md).

Same honesty rule as conflict checking everywhere in this plugin: brand links are corpus-derived, not exhaustive. "Affiliated with X in the Creatorland corpus" ≠ "works with X"; "no affiliation found" ≠ "free of X". Unreported, in-flight, and off-platform deals are invisible. The report says this prominently, not in fine print.

If the user wants a one-shot conflict check on a slate they already have, that's the conflict step inside brief-to-shortlist — this skill is for mapping the landscape AROUND competitor brands. For watching a creator search space over time, use talent-scout (the two compose: a finished competitor-watch search can be saved as a scout profile).

Precision & filters for outreach (convention 14). Broad mapping is right for the landscape layer here — but when the whitespace targets become a slate someone will actually contact, cast that cut for precision: precision: "tight" with the hard-gate filters (platform, niche, data_freshness_days, content_format, audience_country), do not stack dozens of lookalike hops to pad it (each hop drifts), and trim the weak tail by relative_fit (within-set fit, 1.0 = strongest). Canon: ${CLAUDE_PLUGIN_ROOT}/shared/conventions.md convention 14.

Inputs to collect

• Competitor brand names (required, 1–5). Confirm spellings — affinity matching is name-based.
• The user's own brand/vertical context — needed to define "similar creators" for the whitespace layer and to scope the optional market-intel context call. Infer from conversation if possible; ask once if not.
• Geo/platform/tier constraints (optional) — applied to searches if given.
• Whitespace layer on/off — default on; skip if the user only wants the affiliated map.
• Credit mode — default thorough; thrifty on the usual trigger phrases.

Flow

Step 1 — Affinity search per competitor brand. One search_creators call per named brand (convention 9: a brand LIST gets per-brand searches and a per-brand-sectioned deliverable, never one merged search):

search_creators {
  "mode": "brief",
  "brief": "creators creating content around <competitor brand>, <user's vertical> space",
  "filters": {
    "brand_affinities": ["<competitor brand>"],
    "platform": "<if constrained>",
    "country": "<if constrained>",
    "min_followers": <if constrained>
  },
  "limit": 15,
  "precision": "balanced"
}

(thorough: limit 15; thrifty: limit 8.) brand_affinities is the load-bearing filter; the brief text reinforces it.

Step 2 — Profile fan-out for affiliation confirmation. get_creator_profile { "identifier": { "type": "<type from search result>", ... } } on candidates. The profile's brand affiliations field (pro plan) is the confirmation layer: classify each creator as affiliation-confirmed (competitor appears in profile affiliations) vs affinity-signal only (surfaced by the affinity search but not in the affiliations field — weaker evidence, labeled as such). If affiliations are unavailable (free plan), say so: the whole report downgrades to affinity-signal confidence, disclosed up top. Thorough: profile every candidate. Thrifty: top 6 per brand.

Step 3 — Whitespace layer (if on). For each competitor, seed a lookalike search off the 1–2 strongest affiliation-confirmed creators:

search_creators {
  "mode": "lookalike",
  "seed_creator": { "type": "creatorland_user_id", "creatorland_user_id": "<strongest affiliated creator>" },
  "limit": 10,
  "precision": "balanced"
}

Profile the lookalikes, then keep those whose affiliations do NOT include any of the named competitors → the whitespace list. Label honestly: "no competitor link found in the corpus" — never "unaffiliated."

Credit estimate fires here. Thorough with 3 competitors + whitespace runs ~3 searches + ~45 profiles + 3 lookalike searches + ~30 profiles ≈ 87 credits — well over the ~30 threshold, so state the estimate before the fan-out and offer thrifty (proceed after stating; do not block).

Step 4 — Freshness Gate on every profiled creator per the shared module. Affiliations drift; a stale profile's competitor link is exactly the kind of claim that embarrasses a deck. Stale rows go to "re-verify", always.

Step 5 — Optional market context (pro, one call). One query_market_intelligence { "mode": "market", "vertical": "<user's vertical>" } to frame how active the vertical's deal landscape is overall — context for the whitespace pitch. Wrapped in Refusal Recovery (market floor: 5 brands / 25 deals); disclose any broadening. Thrifty: skip.

Step 6 — Write the deliverable.

Deliverable

# Competitor Creator Watch — <competitor brands>
_Built <date> · Creatorland Data · corpus-derived, not exhaustive (see scope note)_

> **Scope note (read first):** brand links below come from Creatorland's
> corpus of known affinities and affiliations. They are not a complete record
> of who works with these brands — unreported, in-flight, and off-platform
> deals are invisible to this data. Treat "affiliated" as "corpus-confirmed
> link" and "whitespace" as "no link found in the corpus", nothing stronger.

## <Competitor brand 1>

### Corpus-linked creators (<N>)
| Creator | Platform / followers | Evidence | Freshness |
|---|---|---|---|
| @<handle> | <platform>, <count> | affiliation-confirmed (profile affiliations) | fresh |
| @<handle> | <platform>, <count> | affinity-signal only | aging (note) |

### Whitespace — similar creators with no corpus link to <brand 1> (<K>)
| Creator | Platform / followers | Similar to | Freshness |
|---|---|---|---|
| @<handle> | <platform>, <count> | lookalike of @<seed> | fresh |

<repeat per competitor brand>

## Cross-brand observations
<creators linked to MORE THAN ONE named competitor; clusters by tier/geo/platform — only patterns the data actually shows>

## Vertical context (if pulled)
<market-mode stat + the provenance line exactly as the tool returned it; if
broadened: "Context basis: <level> — narrower slice was below the privacy
floor (5 brands / 25 deals); a privacy feature of the data source.">

## Re-verify before pitch
<stale rows with what's stale; or "None — all rows cleared the freshness gate.">

## Caveats
- Corpus-derived, not exhaustive — see scope note. "Whitespace" creators may
  have undisclosed competitor relationships.
- <free plan: "Profile affiliations unavailable on this plan — all evidence
  is affinity-signal level," if applicable>
- <thin results per brand, disclosed per convention 8, if applicable>
- No creator contact info is included or available via this tool; route
  outreach through Creatorland connections or creators' public profiles.

---
Data freshness: <N>/<M> creators synced within the last sync window; <K> flagged for re-verification.
Provenance: Creatorland Data MCP · <B> affinity searches + <L> lookalike searches + <P> profiles<+ 1 market context (<clearance level>)> · <date>.
Credits used this run: ~<N> (<B+L> searches ×2 + <P> profiles ×1 <+ market intel ×5>).

Honesty rules

• Corpus-derived, not exhaustive — the scope note leads the deliverable; never present the linked list as a brand's complete roster, never present whitespace as "guaranteed conflict-free."
• Distinguish affiliation-confirmed from affinity-signal evidence on every row; downgrade the whole report's confidence framing on free plan and say why.
• Thin per-brand results: disclose and broaden one labeled step (convention
  8. — never pad a competitor's list with weak fits to look thorough.
• No deal terms, spend figures, or contract claims — the corpus shows relationships, not commercial detail.
• Stale affiliation data never silently mixed in (Freshness Gate).
• No contact info anywhere (convention 7).

Credit footprint

thorough (3 competitors, whitespace on): ~85–95 credits (3 affinity searches ×2 + ~45 profiles + 3 lookalike searches ×2 + ~30 whitespace profiles + 1 market intel ×5; +5 per refusal-ladder rung) — the estimate always fires · thrifty (3 competitors): ~30–35 credits (3 searches ×2 + ~18 profiles + 1 lookalike search ×2 + ~6 profiles, market context skipped) · single-competitor thrifty, whitespace off: ~10 credits.