Insaight — Save Research to Notion

Persist a research brief to Notion so the next research run on the same subject can build on it instead of starting cold.

Scope: research briefs only. The [NOTION_OUTREACH_LOG] page is the user's own record of what they sent, kept by hand — treat it as read-only here. Writing generated messages into it would corrupt the one record of what a real human actually sent.

Step 1 — Verify there's something to save

Look for a structured research brief in the conversation:

• Person: 👤 Role & Context, 🧭 Career Trajectory, …
• Company: 🏢 Company Snapshot, 🎯 Prospect Evaluation, …
• Post: 📌 Post Summary, 💬 Comment Themes, …

If there's no brief, stop and say so. Reconstructing one from memory would save something the research skills never actually produced.

Step 2 — Find the parent page

notion:search(query="[NOTION_RESEARCH_PAGE]")

If it's missing, search for [NOTION_WORKSPACE] (both names come from insaight:get_config()) and create [NOTION_RESEARCH_PAGE] under it — only with the user's confirmation, since creating pages in someone's workspace is not reversible from here. Cache the parent page ID for the session.

Step 3 — Update or create

Search for a page titled with the subject:

notion:search(query="[Name]")

If a page for this exact subject already exists under [NOTION_RESEARCH_PAGE]:

• Update it rather than creating a second page — duplicates split the history the research skills read back on the next run.
• Preserve prior content; replace only the sections the current brief covers.
• Append Last updated: [YYYY-MM-DD].

Otherwise create a new page:

• Person or company: [Name] — [YYYY-MM-DD]
• Post: Post — [author last name] — [topic gist] — [YYYY-MM-DD]

Step 4 — Write the page

Use Notion blocks rather than one markdown blob, so the page stays navigable and individually editable:

• heading_1 for the name/title
• paragraph (italic) for the metadata line
• heading_2 for section headers
• paragraph for body text
• bulleted_list_item for lists
• table for the prospect evaluation rubric

Page structure

# [Name] — Prospect Analysis
*Analysed on [date] | Source: LinkedIn via Insaight*

---

## 👤 Role & Context   (person)
## 🏢 Company Snapshot (company)
## 📌 Post Summary     (post)
...

## 🧭 Career Trajectory        (person)
## 🎓 Education & Credentials  (person)
## 📝 Content Themes
## 🎯 Professional Focus / Prospect Evaluation
## 🔭 Future                   (company)
## 💬 Comment Themes           (post)
## 🔑 Decision-Maker Signals
## 💎 Uncommon Commonalities   (person, if provided)
## 🪝 Outreach Hooks
## ❓ Open Questions

Mirror the sections the brief actually produced — the list above is a superset across brief types, not a template to fill in. Inventing a section puts empty headers in front of the user next time they open the page.

Step 5 — Report and stop

One line:

Saved to Notion: [page title] ([create | update])

Don't re-output the brief. The user is already reading it in the conversation.

Edge cases

• Notion MCP not connected: Tell the user once and let the calling skill's output stand — a missing integration shouldn't swallow finished research. Message: "Notion MCP isn't connected — research completed but not auto-saved. Connect at claude.ai → Settings → Integrations."
• Notion tools not loaded: Run tool_search(query="notion search fetch create") to load the schemas, then retry. Tool names vary between Notion MCP builds, so use the names that come back rather than assuming them.
• Duplicate page titles: Prefer the match under [NOTION_RESEARCH_PAGE]. If it's still ambiguous, ask — writing to the wrong page is harder to undo than asking one question.