Ship the analysis: the CRM update, the post, the email.
Takes a read's output and drafts what goes out, draft-first and fully logged, ready to paste or write directly when connected.
Inside: The operation router, draft-first defaults, and an audit row on every write.
Install it in one line, or paste it in.
~/.claude/skills/ and runs automatically when it is relevant.Connect your context. Set it to your motion.
it writes the field update and the contact record directly, with duplicate checking.
it posts or drafts the message in the right place.
drafts a multi-touch sequence in your voice from a style guide you provide.
This was built for a B2B GTM stack. Set these to your tools:
Wire in one tool or all of them. Each connector unlocks a slice: CRM alone gives you writebacks, a channel tool alone gives you posts.
| Set this | What it is | Default / Example |
|---|---|---|
| CHANNEL tool | where posts and DMs go | |
| OUTREACH tool | where sequences and sends go | your sequencing toolan email client |
| VOICE tool | your playbook and writing style | Octavea pasted style guide |
| DRAFT_FIRST | draft or autosend by default | draft (opt in per workflow to autosend) |
| AUDIT log | where every write is recorded | a log file or sheet you control |
Everything the skill does, in full.
This is the write-side. The other skills read and analyze; this one ships the result. Hand it an account verdict, a usage story, or a ranked list and it drafts what goes out: the CRM field update, the channel post, the personalized follow-up email, the next-touch sequence. Everything is a draft first and everything is logged, so nothing leaves the loop without your eyes on it.
- 1Operation types
One of: crm_field_update, crm_contact_writeback, channel_post, channel_draft, direct_message, outreach_draft, sequence. The skill routes each request to the right operation and returns a structured confirmation.
- 2Draft-first by default
Email drafts and channel messages are created as DRAFTS, not sends, so the person can review before it goes out. Autosend is an explicit, per-workflow opt-in, never the default.
- 3Idempotent CRM writes
Before creating a contact, the skill checks for a duplicate. If one exists, it updates instead of creating, so you never pollute the CRM with duplicate records. Field updates capture the before and after value.
- 4Full auditability
Every write logs a row: timestamp, operation type, target (record and field, or channel and message), before and after values, and which workflow called it. No silent writes, ever.
- 5Honest connector failure
If a step asks to post to a channel but the channel tool is not connected, the skill says so and aborts that operation. It never fabricates a success.
- Draft-first is the default. Autosend requires an explicit opt-in.
- No silent writes. Every operation logs an audit row with before and after values.
- Idempotency on contact writes. Check for a duplicate, update instead of creating.
- Honest failure. A missing connector returns "not connected, operation aborted," never a fake success.
CRM WRITE-BACK COMPLETE · 12 contacts processed Name Operation Notes Contact A CREATED verified email + phone written Contact B UPDATED existed, added mobile + profile Contact C UPDATED existed, updated title Contact D EXISTING (skip) all fields already populated Contact E SKIPPED email not found, not written Summary: 4 created · 6 updated · 1 unchanged · 1 skipped Audit trail: 12 rows appended to your log --- CHANNEL POST (DRAFT) · #gtm-pipeline Mode: draft (autosend off) - review and send when ready Message: THE DAILY DROP - Friday... [full copy ready to post] Audit: 1 row appended
There are no thresholds here, only defaults you control: draft-first is on until you opt a workflow into autosend, and the audit log goes wherever you point it. The one rule that does not move is that nothing sends silently and nothing fakes success. Everything else is yours.
Where an operator takes this next.
The read is step one. Here's where an operator takes it once the manual version proves out.
You built the read once; now it runs itself.
Chain this after closed-lost-analysis or coaching-analyst so every read ends in a real CRM write or Slack post, not a doc nobody acts on.
Opt in field updates and internal Slack posts to autosend while keeping outbound email drafts human-reviewed, so the risky sends still get eyes.
Point the audit log at a Snowflake table so every write across every workflow is queryable in one place.
One skill is the on-ramp.
A single skill does one job. Chained into a playbook, or run as a full build, it becomes a system. Here is where this one plugs in.