Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Job Digest Reporter - Weekly Job Email + Excel

    1

    One weekly email that tells you what is actually new - grouped by role, your shortlist first, spreadsheet attached.

    $9.99

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Job Digest Reporter - Weekly Job Email + Excel

    Job Digest Reporter - Weekly Job Email + Excel

    Example session with this skill installed

    Set up a weekly job digest for platform engineering roles in India, email me Sunday mornings, and attach a spreadsheet.
    
    • Read your context and instructions
    • Compiled the job digest reporter
    $ python3 skills/job-digest-setup/scripts/jobdigest_config.py init --preset india-platform
    Wrote jobdigest.config.json (preset: india-platform)
    
    Next:
      1. Edit roles[] and search.locations for the jobs you actually want.
      2. Add target companies:  python3 jobdigest_config.py add-company "Stripe" --ats greenhouse:stripe
      3. Set email.from / email.to, then export your SMTP password:
         export JOBDIGEST_SMTP_PASSWORD='...'   # the value lives only in your shell, never in this file
      4. python3 jobdigest_config.py validate
    
    $ python3 skills/job-merge-dedupe/scripts/merge_jobs.py examples/jobs.sample.json:sample --out jobs.json --report
    sources:
      examples/jobs.sample.json                11 record(s), 1 skipped (missing company, title or http(s) URL) [sample]
    merged: 10 unique job(s) from 1 source(s), 0 priority
    wrote jobs.json
    
    $ python3 skills/job-digest-state/scripts/digest_state.py diff jobs.json --out digest.json
    state: .jobdigest/seen.json (0 job(s) remembered)
    this run: 10 new, 0 changed, 0 already seen, 2 dropped as older than 7 days
    wrote digest.json
    state NOT updated (pass --commit once the digest has actually gone out)
    
    $ python3 skills/job-digest-excel/scripts/build_workbook.py digest.json --out jobs.xlsx
    wrote jobs.xlsx - 8 job(s), 0 at priority companies, 9521 byte(s)
    
    $ python3 skills/job-digest-email/scripts/send_digest.py digest.json --attach jobs.xlsx --html-out preview.html
    wrote preview.html (13040 bytes)
    
    PREVIEW - nothing was sent and no network was used.
      subject:  Weekly Job Digest - 2026-11-07 (8 new)
      jobs:     8 (8 in the headline count)
      attach:   jobs.xlsx
      would go to: (email.to is empty)
      add --send to deliver it
    

    Read the numbers across that run and you can see the pack's whole argument. Twelve postings in the
    sample file; eleven usable, with one skipped and named as skipped — its "apply" URL was
    javascript:alert(document.domain), which is not a job posting. Ten unique after merging, because
    two of the records were the same Lever posting reached through different tracking links
    (?utm_campaign=weekly&gh_src=abc), folded together once the parameters came off.

    Two dropped as older than seven days, including postings that carried no date at all, aged from the first run that ever saw them.

    Eight in the email, and the headline count is eight because all eight are genuinely new.

    Nothing in that output is a round number, and nothing is silent. Every record that did not make it
    into the email is accounted for on a line that says which stage dropped it and why. That is the
    difference between a digest you trust on week six and one you stop opening.

    Then, once it looks right:

    $ python3 skills/job-digest-email/scripts/send_digest.py digest.json --attach jobs.xlsx --send
    $ python3 skills/job-digest-state/scripts/digest_state.py diff jobs.json --commit
    

    --commit is deliberately a separate step. State is only advanced after the digest has actually
    gone out, so a failed send does not quietly mark eight jobs as "seen" and hide them from you
    forever.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Job boards show you the same forty postings every week. You scroll, you recognise most of them, you are not sure which ones you have already dismissed, and eventually you stop looking.

    This pack is the other half of a job search: the part that remembers. Point it at job data from any source and it merges it, deduplicates it, works out what you have not seen before, and sends you one email — grouped by role, your priority companies first, with a filterable spreadsheet attached.

    Five skills

    SkillWhat it does
    job-digest-setupWrites and validates jobdigest.config.json from one of five presets. Everything else reads that one file.
    job-merge-dedupeMerges any number of job JSON files. Deduplicates on the canonical URL with tracking parameters stripped, so the same role in two cities survives as two jobs instead of one.
    job-digest-stateRemembers what you have been shown, so "new" means new since the last digest rather than "found in this run".
    job-digest-emailThe HTML email: grouped by role, priority companies highlighted, responsive, Outlook-safe. Plus a ranked-shortlist layout via --mode pipeline.
    job-digest-excelA three-sheet workbook — All Jobs, Priority Companies, By Role — with frozen headers, autofilters and clickable Apply cells.

    Working in five minutes

    python3 skills/job-digest-setup/scripts/jobdigest_config.py init --preset minimal
    python3 skills/job-merge-dedupe/scripts/merge_jobs.py examples/jobs.sample.json:sample --out jobs.json --report
    python3 skills/job-digest-state/scripts/digest_state.py diff jobs.json --out digest.json
    python3 skills/job-digest-email/scripts/send_digest.py digest.json --html-out preview.html
    

    Twelve sample postings ship with the pack, so that works before you have configured anything. Preview is the default: nothing is sent and no network is touched unless you add --send.

    Made to be customized

    One file, jobdigest.config.json, in the directory you run from. There is no second place to configure anything.

    You changeIt affects
    roles[]Which titles count as a match, and the section headings you see. One taxonomy, used by every skill.
    companies[] + priorityYour shortlist — its own email section, its own sheet, a highlighted row. aliases[] catches employers that post under more than one name.
    search.locations, max_age_daysWhat gets filtered out. An empty location list means no location filter, not no results.
    brand.*Subject line, title, accent colours, body width.
    excel.sheets.*Sheet names — rename them into your own language if you like.
    email.*Any SMTP provider. Gmail, Outlook, Fastmail, iCloud, Amazon SES, Resend and Postmark are all documented in shared/smtp-providers.md.
    scoring.bands, urgent_daysScore thresholds and the closing-soon cutoff in --mode pipeline.

    Five presets ship as starting points — minimal, us-remote-swe, india-platform, eu-data, uk-finance — and JOBDIGEST_* environment variables override anything per machine without editing the file.

    Your mail password never enters the config. email.password_env holds the name of an environment variable, which is what makes the file safe to commit to a repo. validate rejects any secret-looking key under email rather than trusting you to remember.

    Any source of jobs

    A valid input is a JSON list of records with a title, a company and an http(s) url. Everything else — location, date, work mode, skills, your own scores — is optional enrichment. Field aliases (jobUrl, applyUrl, postedAt, companyName, …) are accepted and normalized, so an export from another tool usually works unchanged; skills/job-merge-dedupe/reference/adapters.md has worked adapters for half a dozen shapes. If you want the pack to go and get the jobs too, that is the companion Job Source Fetcher pack.

    Treats scraped text as data, never as instructions

    Every interpolated field is HTML-escaped, every URL is scheme-checked, and a javascript: or data: URL renders as plain text with no link and gets no spreadsheet hyperlink. Titles starting =, +, - or @ are neutered so Excel does not evaluate them as formulas. Job data comes from pages other people control, and this pack is written on that assumption throughout.

    Compatibility

    EnvironmentSupport
    Claude Code · claude.ai · Claude Agent SDKSkills auto-discovered from skills/
    Any other coding agent (Cursor, Windsurf, Copilot, Codex, Gemini CLI, …)Works — point the agent at USAGE.md or the relevant SKILL.md
    Plain terminal, no agent at allWorks — USAGE.md is copy-paste

    Python 3.10+, standard library only except an optional openpyxl for the spreadsheet. macOS, Linux, Windows. No Node, no Bun, no API keys, no accounts, no telemetry, no phone-home. Mail goes through your own SMTP account.

    Includes MIT-licensed portions of ai-job-search (© 2026 Mads Lorentzen); modifications and new work © 2026 Sujal Goswami. Full notice: shared/LICENSE-MIT.md. Not affiliated with or endorsed by ai-job-search.

    • This pack reports; it does not fetch. It needs job data as input — from the companion
      Job Source Fetcher pack, from your own script, or from any tool that can emit JSON. The
      sample file lets you evaluate everything before you wire up a source.
    • No LinkedIn, Indeed or Glassdoor scraping, and there never will be. Their terms forbid it
      and automating it risks your account, not ours.
    • The spreadsheet needs openpyxl (pip install openpyxl). Without it that one skill exits
      with a clear install hint instead of failing halfway through a scheduled run; the email works
      regardless.
    • Mail goes through your own SMTP account. No sending service is included, and Gmail needs an
      app password rather than your login password.
    • No ranking of which jobs matter most to you. Priority is a flag you set per company, not
      a model's opinion.
    • Dateless postings are aged from when the pack first saw them, not from when they were
      posted, because many sources publish no date. A posting that was already three months old on
      first sight counts as new that week.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 9 days ago

    Requires

    Gmail
    Google Account
    Python
    Git

    Permissions

    Terminal / Shell
    Network Access
    Read Files
    Write Files
    Environment Variables

    Allowed Hosts

    gmail.googleapis.com
    accounts.google.com
    Listed9 days ago

    What's inside

    Frequently Asked Questions

    The Job Source Fetcher provides pre-built logic for five major ATS platforms (Greenhouse, Lever, Ashby, Workable, SmartRecruiters) and custom career pages, saving you from writing and maintaining individual scrapers for every company.

    This skill requires a Python 3.10+ environment and can be integrated into any AI agent capable of executing Python scripts or managing GitHub Actions and cron jobs.

    You will receive the full Python source code for ATS fetching and career page scanning, example configuration files, and automated scheduling templates for GitHub Actions and Windows Task Scheduler.

    The skill includes built-in 'loud-fail' error reporting that alerts you if a specific company's page structure changes, ensuring you don't miss roles due to silent script failures.

    No, this tool is designed for maximum portability and zero-dependency overhead, using only Python's standard library to avoid complex installation processes.

    This skill does not execute JavaScript; it is optimized for speed and safety by reading public JSON endpoints and static HTML, meaning it will not work on sites that require client-side rendering.

    New in Workflow Automation