| name | counter-pitch |
| description | Draft thoughtful rebuttal email when classifier returns objection_with_counter |
Counter-Pitch Skill
When the reply-classifier (skills/email-automation/scripts/reply-classifier.py) labels an inbound message objection_with_counter — the prospect raised a concern but stayed engaged — Rick drafts a counter-pitch reply that:
- Addresses the SPECIFIC objection (not generic boilerplate)
- Cites a relevant case study when one matches the objection theme
- Soft re-pitch with a new angle — not "but you should buy because…", more "here's what I'd actually do for you"
- Founder voice: dry humor, direct, no buzzwords, opinion-first
- Short: 5-7 sentences max. ~200 words.
Inputs (CLI args)
--thread-id — the email_threads.thread_id we're replying to--objection-text — the actual objection text from the classified row--prospect-id — optional, looks up prior context if present--dry-run — print the draft without writing to disk
Output
JSON file at ~/rick-vault/mailbox/drafts/counter-pitch/<thread_id>-<ts>.json:
{
"draft_id": "cp_xxxxxx",
"thread_id": "...",
"subject": "Re: …",
"body": "…the drafted reply…",
"objection_class": "pricing|timing|fit|trust|other",
"case_study_cited": "newton.md or null",
"model": "claude-sonnet-4-6",
"created_at": "ISO timestamp"
}
Safety rails
- NEVER auto-sends — there is no Resend/SMTP code in this skill at all. Drafts only.
- Vlad reviews via /inbox Telegram (TIER-3.5 #A12, separate ship). Until that ships, drafts pile up in the directory and Vlad reviews them on the file system.
RICK_COUNTERPITCH_AUTOSEND env — DECORATIVE in this skill (no send code path exists to gate). Reserved for a future sender daemon. Either value (0 or 1) writes drafts only.- Cost budget: ~$0.05-$0.15 per draft (claude-sonnet-4-6 via writing route). Falls under TIER-0 #3
content_gen workflow budget cap if dispatched as a workflow.
Voice rules (reinforce in prompt)
- Cite the source URL of the case study if used: "Newton (callsign of our $9 customer) had the same concern about X — turned out…"
- Never claim metrics that aren't in our actual data. Real MRR is $9, 1 paying customer. Per SELF-FAQ.
- End with ONE concrete next step (15-min call OR specific question OR offer to share something useful).
- Em dashes OK. NEVER use "I hope this finds you well" or "I wanted to follow up regarding".
