Every role is scored on five dimensions. Dimensions 1 and 2 are primary gates— if a role fails either of them, it doesn't pass regardless of other scores. Dimensions 3–5 add nuance and drive the final score.
| Dimension | What It Measures | Pass Criteria (customize for your search) | Weight |
|---|
| D1 — Role Scope | Does this role match the type of work you want to do? | Must be [your target scope, e.g., upstream strategy / positioning / GTM] — not pure execution or management of a single channel | Primary gate |
| D2 — Industry Fit | Is this the industry / sector you want? | Tier 1: [your preferred industries]. Tier 2: [acceptable-if-inbound industries]. Auto-fail: [industries you won't consider] | Primary gate |
| D3 — Competitive Position | Are you a competitive candidate for this role? | Strong = your background is a direct match. Moderate = adjacent. Weak = stretch. | High |
| D4 — Compensation | Does the comp range meet your floor? | Minimum acceptable: $[X]. Auto-fail if explicitly below floor and no path to it. | Medium |
| D5 — Structure / Never Again | Any structural red flags? | Auto-fail: [company types, structures, or role characteristics you'll never accept, e.g., pure agency work, no remote, contract-only, etc.] | Veto power |
Scoring Guide
| Score | Meaning | Action |
|---|
| 9–10 | Exceptional match — rare | Apply immediately. Prioritize above all else. |
| 7–8 | Strong match — most primary criteria met | Apply. Move to outreach or application within the week. |
| 5–6 | Moderate match — needs more information | Pull the full job description. Re-score before deciding. |
| 3–4 | Weak match — one or more primary criteria fail | Log and watch. Do not invest time unless circumstances change. |
| 1–2 | Clear mismatch | Log and close. Do not pursue. |
Key principle
Score based on what you can see in the job alert or ATS listing. If a role sounds potentially right but the listing is thin, mark it 5–6 MONITOR and pull the full JD before re-scoring. Don't commit to applying or ignoring based on incomplete information.
07Running the Morning Sweep
The sweep is a structured workflow the AI agent follows each time it runs. You give Claude a detailed instruction document (saved as a text or markdown file in your job search folder) that tells it exactly what to do.
What the Sweep Instruction Document Should Cover
- Your identity and search context — your target role, seniority level, industries, and any absolute deal-breakers. This is how Claude knows which roles are relevant.
- Gmail sweep instructions — which sender addresses to search, how far back to look, which inbox tabs to include.
- ATS sweep instructions — read
Target_Companies.csv, navigate to each ATS URL, search for relevant role keywords, extract new listings. - Deduplication rule — check
job_log_MASTER.xlsx before logging anything. Skip if already logged. - Scoring instructions — apply the 5-dimension framework to each new role found.
- Logging instructions — add new rows to
job_log_MASTER.xlsx; update Last Checked in Target_Companies.csv. - Report format — how to structure the morning sweep report (see below).
Sweep Report Format
At the end of each sweep, Claude generates a structured markdown report. The report should include:
- Sweep stats: Number of companies checked, Gmail queries run, new roles found, total log rows
- New roles by score tier: GREEN (apply), YELLOW (monitor), RED (skip) — with score, verdict, and brief rationale for each
- Top picks: The 3–5 roles that need attention first and why
- Previously logged roles still live: High-value roles already in your pipeline confirmed still open
- Access failures: Any companies that couldn't be checked and why
- ATS issues to fix: Broken URLs or tokens to update before the next sweep
Typical Sweep Sequence (What Claude Does)
- Read
Target_Companies.csv to get the full company list and ATS URLs - Read
job_log_MASTER.xlsx to load already-logged roles for deduplication - Search Gmail for each configured alert sender — extract role titles and companies from email content
- For each company in Target Companies: navigate to ATS URL, search for relevant roles, extract new listings
- For each new role: apply 5-dimension scoring, assign a verdict
- Log all new roles to
job_log_MASTER.xlsx - Update Last Checked date for each company in
Target_Companies.csv - Generate and save the morning sweep report
Important
Some ATS platforms (iCIMS, Workday, some custom sites) require JavaScript to render and cannot be read by automated agents. Others use bot detection. For these, note them in your Target Companies list as "check manually." Expect 5–15% of boards to need manual handling at any given time.
08Automating the Sweep with Claude Cowork
Claude Cowork supports scheduled tasks — you can set the sweep to run automatically on a recurring schedule so it happens whether or not you remember to start it.
Setting Up a Scheduled Sweep
- Open Claude Cowork and navigate to your job search folder
- Ask Claude to create a scheduled task: "Run my job search sweep every Monday morning at [time]"
- Point the scheduled task at your sweep instruction file
- Verify the schedule is set correctly and confirm the first test run
The sweep will run in the background on your schedule. When it's done, the morning sweep report will appear in your job search folder, ready for review when you open your laptop.
Recommended Sweep Frequency
- Monday morning — covers all weekend postings from Friday–Monday (highest volume)
- Wednesday or Thursday — catches mid-week postings before Friday's slowdown
- On-demand — run any time you want a fresh check, especially if you're actively interviewing
Two sweeps per week is a good rhythm for an active search. Once per week is acceptable for a passive search.
09Tracking Your Pipeline
Your job log doubles as your pipeline tracker. Use the Status column to move roles through stages as you take action.
| Stage | Meaning | Trigger |
|---|
| New | Scored and logged, no action taken yet | Automatically assigned at logging |
| Reviewing JD | Full job description pulled, re-scoring in progress | When you pull the full JD for a YELLOW role |
| Outreach Sent | You've reached out to someone at the company | After sending a LinkedIn message or email |
| Applied | Application submitted | After clicking submit |
| Recruiter Screen | Scheduled or completed initial recruiter/HR screen | After scheduling |
| HM Interview | Scheduled or completed hiring manager interview | After scheduling |
| Panel / Final | In final interview stages | After scheduling |
| Offer | Offer received | When offer arrives |
| Rejected | No longer in process | When rejected or withdrew |
| Closed | Role no longer available or you've chosen to skip | When role is removed or you pass |
Follow-Up Cadence
Use the Next Action and Last Action Date columns to drive follow-up. A simple rule of thumb:
- After outreach: follow up in 5–7 business days if no response
- After application: follow up in 7–10 business days if no response
- After interview: follow up within 24 hours with a thank-you note
- If ghosted after two follow-ups: move to Closed, keep the company on your radar
10Continuous Improvement
The system gets better over time as you fix ATS URLs, tune your scoring criteria, and refine your target list. After each sweep, review:
- Access failures — fix broken ATS URLs before the next sweep. Each broken URL is a company you're not monitoring.
- Score calibration — if you're getting too many YELLOWs that never turn into applications, tighten your D1/D2 criteria. If you're getting too few results, check whether your keyword search is too narrow.
- Company list health — add companies you discover during the sweep. Remove companies that have been dormant for 3+ months or that have clearly stopped hiring in your function.
- Gmail alert coverage — if a role from a target company showed up somewhere unexpected, add that source to your alert setup.
Key principle
What makes this system work: Consistency.The value compounds over time — the longer you run it, the more comprehensive your deduplication becomes, the better you get at scoring, and the cleaner your company list gets. A sweep run once a week for four weeks will surface patterns you'd never catch manually.
This guide covers a generic implementation of the automated job search monitoring system. Adapt the scoring framework, company list, and sweep schedule to your specific search context.
