Enterprise clients with co-marketing obligations — Intel, Qualcomm, Snapdragon — face million-dollar penalties for wrong CPU logos, incorrect brand stickers, or competitor UI visible in mockups, and currently have no automated check before assets are delivered. · Seen in: 2026-06-04-1259-hande-maximilian-magier-FlzO

A Claude agent that reads a delivery brief, extracts the named co-marketing partners and their compliance rules, and runs a structured checklist against the asset description before delivery — flagging any rule that cannot be confirmed as met and blocking delivery until the PM signs off on each flag.

Build a /compliance-check skill that accepts a brief basename and a client compliance profile JSON. The profile encodes the partner rules Hande described: correct CPU logo variant per product line, no competitor UI in screen mockups, correct brand sticker placement, NDA asset confidentiality confirmation. The agent outputs a checklist markdown with pass/fail per rule and a required PM sign-off field. Wire it as a required step in the delivery workflow for any client tagged `enterprise-compliance` in the client wiki. The ASUS H2 retainer is the first pilot — Hande's QBR pitch on Friday is the moment to name this capability as a differentiator.

Hande named brand compliance penalties as a primary pain and NDA-backed confidentiality as a hard requirement for switching agencies — leading with this in the proposal this week is the fastest path to closing the retainer before a competitor does.

Subscription clients like Blume Health progressively expand their requests to include strategy consultation, content generation, and daily high-touch updates that are outside the standard scope, and PMs have no automated signal that a client is drifting out of bounds until the margin is already negative. · Seen in: 2026-06-04-1200-magier-weekly-J5Cc, 2026-06-03-0859-domains-alignment-surY

An n8n workflow that reads new tickets from Many Requests, classifies each request against a scope definition JSON for that client's subscription tier, and posts a Slack alert to the PM and Anna when a ticket contains scope-exceeding keywords — strategy, content creation, daily update, consultation — before work begins.

Build an n8n flow triggered by the Many Requests webhook on new ticket creation. The flow sends the ticket description to Claude with a system prompt encoding the S/M/L tier scope boundaries. Claude returns a classification: in-scope, borderline, or out-of-scope, with a one-line reason. Out-of-scope tickets trigger a Slack message to the PM channel with the ticket link, the classification reason, and two action buttons: escalate to Anna or reply to client with scope note. Blume Health is the first pilot — their ticket history from the past month is the training set for calibrating the borderline threshold. The scope definition JSON is owned by Anna and updated when special deals are signed.

Blume Health's fourth pipeline starts Monday and the margin impact of untracked out-of-scope work is already visible in the May P&L — every week without this signal is another week of absorbed cost.

The team is running content experiments on LinkedIn to recover from a 95% reach collapse but has no structured way to track which post formats, topics, and tagging patterns are producing recovery signals versus continuing to decay. · Seen in: 2026-06-04-1200-magier-weekly-J5Cc, 2026-06-04-0700-daily-stand-up-Egsv

A Claude agent that reads a weekly LinkedIn performance export, scores each post against a format taxonomy — static vs video, personal vs educational vs team, tagged vs untagged — and outputs a ranked signal table showing which format-topic combinations are recovering and which are still declining, with a recommended content mix for the following week.

Build a /linkedin-signal skill that accepts a CSV export from LinkedIn analytics and a post log markdown that Gabriel or Max maintains with one line per post: date, format, topic category, tagged accounts. The agent joins the two datasets, computes impression-per-post by category, and outputs a markdown table ranked by recovery rate. The output feeds directly into the weekly content brief. The first run uses the past four weeks of data Max already has access to in LinkedIn analytics. The post log format is a three-field append-only file: `[YYYY-MM-DD] [format] [topic]` — low friction to maintain.

The webinar OKR at 259/500 depends on LinkedIn distribution, and the team is currently flying blind on which content is recovering — without a signal tracker, the experiment will run for weeks before anyone knows if it is working.