name: cold-outreach-board
description: Review a cold outreach email (sales, partnership, guest post, investor, or job outreach) for the three failure modes that get it ignored or filtered — spam-trigger language and formatting that hurts deliverability, tone-deaf framing that reads as "about me" instead of "about you," and a response rate that's realistically near zero because of a weak or missing ask. Three independent reviewer lenses each flag issues blind, then cross-examine each other, then a board verdict tags every finding CONFIRMED/CORROBORATED/DISPUTED. Use when asked to review a cold email or outreach message before sending.
trigger: The user pastes or references a cold outreach email and asks for a review, feedback, or "will this get a reply" — or asks specifically about spam filters, tone, or response rate for outreach copy.
version: 1.0.0
compatible_agents: [claude-code, codex, openclaw]

Cold Outreach Board — independent review, then cross-examination, then a verdict

This is a three-phase board process, not three opinions dumped side by side. Phase 1 finds
candidate issues independently; Phase 2 makes the board argue over them to kill false positives
and confirm real ones; Phase 3 delivers one verdict the sender can actually act on before hitting
send.

Anti-sycophancy rule (applies throughout)

Do not default to encouragement. Your job is to find the specific reasons this email gets deleted,
filtered, or ignored — not to affirm that it's "polite" or "professional." Ground every finding in
the actual email text shown to you; never invent context about the recipient that isn't given. If
no recipient/context was provided, note that personalization findings are best-effort and ask for
who this is going to for a sharper pass.

Phase 1 — Three independent passes (blind)

Run each lens as an independent pass. Each lens must reach its own conclusion without seeing the
others' output yet. For each finding, capture internally: what's wrong, the concrete consequence
(lands in spam, gets deleted in 2 seconds, no clear reason to reply) — not a vague warning — and,
if quick, the one-line fix. Quote the actual email line.

🚫 Spam Filter Simulator

Lens: would this actually land in the inbox? Look for classic spam-trigger words/phrases ("free,"
"guarantee," excessive urgency, ALL CAPS or excessive punctuation), a subject line that reads as
clickbait or mismatched from the body, missing personalization that makes it look like a mail-merge
blast, and link/formatting patterns (too many links, tracking-heavy URLs, attachments from an
unknown sender) that trip filters. Ask "does anything here look like what a spam filter is trained
to catch?"

🎭 Recipient Empathy Check

Lens: does this read as "about me" or "about you"? Look for opening lines that lead with the
sender's product/company instead of a reason specific to this recipient, no evidence the sender did
any homework on the recipient (their work, their company, a recent post/launch), asks that would
take real effort from someone who has never heard of the sender, and a tone mismatch for the
relationship (too familiar for a cold intro, or too stiff/salesy for the context). Ask "if I
received this with zero context, would I feel understood or targeted?"

📊 Response Rate Realist

Lens: is there an actual reason to reply, and is it easy to say yes to? Look for no clear single
ask (or too many asks at once), an ask that requires too much effort for a first touch (a 30-minute
call vs. a 2-minute reply), no specific reason "why now" or "why you," and a message so long the
core ask is buried. Ask "what's the smallest possible 'yes' this email is asking for, and is it
obvious in the first two lines?"

If a lens finds nothing after genuinely looking, note that plainly — but that should be rare on a
first draft.

Phase 2 — Cross-examination (the board argues)

Now show each lens the other two lenses' Phase 1 findings (never its own). For every finding from
another lens, each reviewer answers two questions in one line each:

1. Challenge: "Is this actually a risk given the full email, or is it covered elsewhere /
   overstated?" If a reviewer can name a concrete reason a finding is a false positive (e.g. the
   personalization is actually there in paragraph two, or the ask is already narrowly scoped), the
   finding gets challenged.
2. Corroborate: "Does this finding also matter through my lens?" (e.g. a generic opening the
   Recipient Empathy Check flagged might also be the exact mail-merge signal the Spam Filter
   Simulator already noted — corroboration from a second, independent lens is a stronger signal
   than one lens's opinion alone.)

Do not let a lens skip this step or rubber-stamp everything as fine.

Phase 3 — Board verdict

Merge everything into one report. Tag every surviving finding:

• CONFIRMED — no lens challenged it, or a challenge was raised but not substantiated with a
  concrete reason.
• CORROBORATED — flagged independently or corroborated by more than one lens. Treat these as
  the highest-priority findings in the report.
• DISPUTED — a lens raised a concrete, specific reason to doubt it. Include the finding and
  the challenge side by side so the sender makes the final call, rather than silently dropping it.
• Drop nothing silently. A finding that was successfully challenged still gets listed, marked
  DISPUTED.

End with one Send recommendation line: e.g. "Rewrite first — 2 CORROBORATED findings (buried
ask, generic opening), fix those before sending" or "Send after fixing 1 CONFIRMED spam-trigger
phrase" or "Send as-is — no CORROBORATED or CONFIRMED blockers, DISPUTED items are informational
only."

Output format

## Phase 1 — Independent findings
### 🚫 Spam Filter Simulator
- ...
### 🎭 Recipient Empathy Check
- ...
### 📊 Response Rate Realist
- ...

## Phase 2 — Cross-examination
- [Spam Filter #1] challenged by Recipient Empathy: "..." / corroborated by Response Rate: "..."
- ...

## Phase 3 — Board verdict
1. **[CORROBORATED]** ... — fix: ...
2. **[CONFIRMED]** ... — fix: ...
3. **[DISPUTED]** ... — challenge: "..." (your call)

**Send recommendation:** ...

Handling sensitive info in the outreach

If the email references real names, companies, or contact details, treat that as context only — do
not echo it back beyond what's needed to explain a finding. This Skill sends the text you paste
as-is to the model; redact anything you'd rather not share before pasting.