Notion Opportunity Database

Purpose

Use this skill when a Notion database should act as the system of record for opportunity workflows.

The database owns discovery, deduplication, scoring, rationale, application handoff, and post-action status. It should feed one selected opportunity at a time into execution skills such as job-application-operator.

Tool Boundary

Use Notion MCP/API tools for:

• fetching database schema
• querying candidate rows
• checking duplicates
• creating rows
• updating row properties
• changing schema and views
• writing application status back

Do not use browser automation to edit Notion if MCP/API access is available.

MCP Query Limitation

Some Notion MCP environments advertise a SQL/data-source query tool but reject it at runtime with an error like Tool notion-query-data-sources not found. When that happens, do not block on full-table SQL scans.

Use these workarounds instead:

1. Create or use a filtered Notion view for the exact audit queue, such as Status is empty, Status is Needs enrichment, or Status is To do.
2. Fetch/search rows from that visible queue and update individual pages through MCP update_page.
3. If view-based enumeration is unavailable, use Notion search within the data source with broad role/company terms to discover likely rows, then fetch/update individual pages.
4. For a known dataset, use previously collected page IDs as a bounded backfill list.
5. If complete reliability is required and credentials are available, use the official Notion API database-query endpoint from a local script instead of the broken MCP SQL wrapper.

Prefer a temporary audit view over search when possible because it has a visible completion condition: the view becomes empty.

Recommended Schema

Minimum fields:

• Role or Name: title
• URL: URL
• Company: text
• Status: select
• Priority Score: number, 0-100
• Rationale: rich text
• Role Family: select
• Location: multi-select
• Salary Range: text
• Salary Rationale: rich text
• Source: select

URL should be one of the first visible columns, ideally immediately after the title.

Status Semantics

Status is the application/execution state. It is not a priority or fit label.

Recommended values:

• Needs enrichment: found or imported, but fit, score, rationale, salary, or source fields are incomplete
• Enriching: an agent is researching and completing the row
• Backlog: enriched and scored, not yet selected for application work
• To do: selected from the backlog for the next application work batch
• In progress: application form is being inspected, filled, or prepared
• Human blocked: requires user action, login, CAPTCHA, missing information, unknown required field, or a policy decision
• Ready to apply: application is prepared and waiting for final submission/review
• Submitted: application has been submitted
• Waiting for response: submitted and waiting for company response
• Interview invite: company invited the candidate to book an interview, but it is not scheduled yet
• Interview scheduled: interview is booked on the calendar or otherwise confirmed
• Interview completed: interview happened and the candidate is waiting for the next response
• Assessment received: company sent an online assessment/challenge, but work has not started
• Assessment in progress: assessment/challenge is actively being worked
• Assessment submitted: assessment/challenge was submitted and the candidate is waiting for the next response
• Rejected: rejected or self-rejected after application
• Closed: job is no longer open

Use Priority Score for ranking and Rationale for fit explanation.

Expected flow:

Needs enrichment -> Enriching -> Backlog -> To do -> In progress -> Ready to apply -> Submitted -> Waiting for response

Human blocked is a branch from In progress; return to In progress or Ready to apply after the blocker is resolved.

Company responses branch after submission:

Waiting for response -> Interview invite -> Interview scheduled -> Interview completed -> Waiting for response
Waiting for response -> Assessment received -> Assessment in progress -> Assessment submitted -> Waiting for response
Waiting for response -> Rejected

Enrichment Criteria

Full enrichment requires candidate context. Before scoring fit or writing a candidate-specific rationale, confirm access to:

• resume or experience evidence
• target role preferences
• location/remote constraints
• compensation expectations, if relevant
• application strategy or dealbreakers

If that context is unavailable, ask whether to do partial enrichment instead. Partial enrichment may still fill public/company facts such as URL, company, role family, location, source, salary range, and salary rationale, but should leave Priority Score and candidate-fit Rationale blank or explicitly mark them as needing candidate context.

A row is enriched enough for Backlog when it has:

• valid URL
• Company
• Role Family
• Location
• numeric Priority Score
• consolidated Rationale
• Source
• Salary Range when a credible value is available, otherwise empty
• Salary Rationale explaining source, confidence, or why salary is unknown

Use Needs enrichment for raw imports and partially captured roles. Use Enriching while an agent is actively researching the row. Do not move a row to Backlog for fit-based application selection until full enrichment has candidate context, unless the user explicitly accepts partial enrichment. Do not move a row to To do until it is in Backlog unless the user explicitly chooses to apply without full enrichment.

Rationale Field

Keep one consolidated Rationale field instead of separate columns for:

• fit rationale
• experience signal
• application angle
• main risk

Recommended structure:

Fit: ...
Experience signal: ...
Application angle: ...
Main risk: ...

Salary Fields

Keep the value separate from the reasoning:

• Salary Range: only the actual range or value, such as EUR 80k-100k base or empty if unknown
• Salary Rationale: source, confidence, or why the value is unknown

Do not put sentences like not captured; verify directly in Salary Range.

One-At-A-Time Application Handoff

Before invoking an execution skill:

1. Fetch the database schema.
2. Select one row by status and score, normally highest Priority Score where Status is To do.
3. Confirm the row has a valid URL.
4. Update Status to In progress.
5. Pass a compact handoff object to job-application-operator.

Handoff object:

source: notion
pageId:
role:
company:
applicationUrl:
priorityScore:
rationale:
salaryRange:
salaryRationale:
roleFamily:
location:
statusBefore:

The application operator should work on exactly that one job, then return a structured outcome.

Writeback After Application

After execution, update the Notion row:

• Status: Ready to apply, Submitted, Human blocked, Closed, or another application-stage value
• Rationale: append concise application outcome notes if they materially change the decision
• Salary Range and Salary Rationale: update only if the application revealed reliable compensation information

Do not overwrite source rationale with application logs. Keep detailed field inventories and sensitive application details in private logs.

Discovery Insert Rules

Before creating a row, deduplicate by:

• exact URL
• normalized URL domain/path
• company + role title

If a duplicate exists, update missing fields and leave application status intact unless the job is closed or reopened.

Output

Return:

• database used
• selected row
• handoff object
• status before and after
• fields updated
• blockers