notion-application-sync - SKILL.md Agent Skill

name: notion-application-sync description: Optionally mirror Liam Van's application tracker data into Notion from the generated visualizer JSON cache. Use only when the user explicitly asks to sync, audit, configure, or schedule Notion matching; normal recruiting skills should not call this by default.

Notion Application Sync

Use this skill only for the optional Notion mirror. The recruiting source of truth remains:

application-trackers/applications.md
application-visualizer/src/data/tracker-data.json
Notion, as a slower external mirror

Do not run this inside normal resume tailoring, Gmail refresh, LinkedIn outreach, or company prospecting unless the user explicitly asks for Notion.

Commands

Preview matches without changing Notion:

python3 skills/notion-application-sync/scripts/sync_applications_to_notion.py --dry-run

Sync from the generated website data cache:

NOTION_TOKEN=secret_... python3 skills/notion-application-sync/scripts/sync_applications_to_notion.py --update-title

Sync one row:

NOTION_TOKEN=secret_... python3 skills/notion-application-sync/scripts/sync_applications_to_notion.py \
  --posting-key "posting-id-or-ats-id"

Refresh the website cache before a sync when tracker files changed:

python3 skills/application-visualizer-refresh/scripts/refresh_visualizer_data.py

Scheduled Automation

The repo includes .github/workflows/notion-application-sync.yml, which can run every 12 hours.

Required GitHub secret:

NOTION_TOKEN: Notion internal integration token shared with the configured tracker database

Optional repository variables:

NOTION_SYNC_ENABLED=true: required before the scheduled job will push to Notion

The job uses application-trackers/notion-config.md for the Notion database/data source and reads application-visualizer/src/data/tracker-data.json as the normalized tracker cache.

Matching Rules

Match Notion rows in this order:

exact Posting Key
exact normalized Company plus Role

If there is no unique match, skip the row and report it. Do not create duplicates automatically.

Guardrails

Markdown and visualizer JSON are authoritative; Notion is not.
Do not sync without an explicit user request or the scheduled workflow being enabled.
Do not guess Notion matches when multiple pages are plausible.
Use --dry-run before a large manual sync.
Keep this optional so the fast recruiting loop is not blocked by Notion API latency.