Skip to content
Conduit

skill

trajecktory

Run your entire job search from one local, AI-powered dashboard: find the roles worth your time, tailor for each, work the right people, and keep every follow-up alive. Open source, your data stays yo

Set up

Skills are installed by placing their folder where your agent loads skills.

The download comes from github.com/michaelinghilterra-creator/trajecktory itself, not from our server, so what you get is exactly what its author published.

Install instructions come from the README

Skills do not use an MCP config. Follow the steps in the README below.

README

synced Oct 17, 2026

trajecktory


Why this exists

A modern job search is a dozen disconnected tools: a spreadsheet, forty browser tabs, your inbox, LinkedIn, a folder of resume versions, and a nagging feeling you forgot to follow up with someone. The busywork crowds out the two things that actually move a search forward: reaching the right people, and putting a genuinely tailored application in front of them.

trajecktory collapses all of it into one local command center. You control your entire job-search ecosystem and workflow from a single dashboard on your own machine, so your time goes to the few roles worth it and the people who can actually help.

What it does for you

  • Saves you time on the wrong roles. It scans job boards (Greenhouse, Ashby, Lever, company pages), scores every posting against your actual CV with a structured A-to-G report, and tells you which handful is worth your energy. It recommends against anything below a real fit, because your time and the recruiter's are both worth respecting.
  • Tailors each application, for real. One click produces an ATS-clean Word resume and cover letter tuned to that specific posting, down to reordering your bullets to lead with what the role cares about. Tailoring is the single largest lever in the data, and this is the lever pulled.
  • Makes sure nothing goes cold. It schedules your follow-ups on a sensible cadence, shows you at a glance who is overdue and who to give up on, and drafts the nudge for you. The thread you would have forgotten is the one it surfaces.
  • Finds and works the right contacts. An in-network (target-talent) CRM, a LinkedIn connection queue for people you can only reach there, and LinkedIn and X posts drafted in your voice and scheduled through Buffer, so you build inbound while you work outbound.
  • Reads your inbox so you do not have to. An optional, read-only Gmail connection catches replies and bounces, logs each against the right application, and drafts your response. It never sends anything on its own.
  • Every AI-written message is yours to edit. Outreach emails, LinkedIn notes, follow-ups, posts: each one lands in an editable field, and you tweak it before it goes anywhere. Nothing is a black box you have to accept.
  • Coaches you with honest numbers. A weekly scorecard measures the inputs you control against floors and your outcomes against sourced benchmarks. Conversion is reported by the furthest stage each role ever reached, thin samples are flagged as too few to rate rather than guessed, and there are no fabricated benchmarks anywhere.
  • Preps you for the room. Per-company interview prep opens as a glance (who, when, the opening, the one story, the reset) with the deep reference one click away, not a wall of text.

Not a spray-and-pray tool. trajecktory is a filter, not a firehose. It surfaces the few roles worth applying to and argues against the rest. It never submits anything: it fills the forms, drafts the answers, and generates the documents, then stops so you make the final call.

Local, private, yours

Everything runs on http://localhost:3333, bound to 127.0.0.1. Your CV, contacts, tracker, and mail stay on your machine and are sent only to the AI provider you choose. No cloud, no database, no telemetry, no account. The tool works for you and answers to you.

Getting Started

Two ways to run trajecktory. Both need one thing and one thing only: a paid Claude plan (Pro or Max). Everything else is either bundled for you or a single copy-paste block below.

Option 1: Windows installer (easiest, nothing to set up)

  1. Download the installer from the latest release. Its name starts trajecktory-setup- and ends in the version number, so grab the newest one. It bundles Node, Chromium, Claude Code, and Git, so you install nothing else.
  2. Run it (a few clicks). It is not code-signed yet, so Windows SmartScreen may warn "unknown publisher": click More info -> Run anyway. If it asks you to restart, do it (that puts Git on your PATH, which Claude Code needs).
  3. Open trajecktory from the desktop or Start Menu shortcut. It opens at http://localhost:3333.
  4. Sign in to Claude once, then work the in-app Launchpad (it walks you through your CV, profile, and target companies with a readiness meter).

New to all of this? The three illustrated PDF guides, attached to every release, walk you through it screen by screen: Setting up Claude, Installing trajecktory, and Using trajecktory day to day. The same day-to-day guide is built into the app under Setup -> Day-to-day guide.

Option 2: Run from source (macOS / Linux / Windows, for developers)

Prerequisites: Node.js 24.21 or newer (the version CI and the bundled runtime use), Git, and Claude Code installed and signed in (claude login).

# 1. Clone and install
git clone https://github.com/michaelinghilterra-creator/trajecktory.git
cd trajecktory
git config core.hooksPath .githooks     # enable the commit guards (recommended)
npm ci                                  # root dependencies
npm --prefix dashboard-web ci           # dashboard dependencies
npx playwright install chromium         # liveness checks + portal scraping

# 2. Launch the dashboard
npm --prefix dashboard-web start        # -> http://localhost:3333

Open the dashboard and the Launchpad walks you through adding your CV, profile, and target companies. Run node doctor.mjs anytime to validate prerequisites. Fuller from-source notes: docs/SETUP.md.

Credentials: Evaluate, Scan, and every AI writing draft run on your own Claude Pro/Max login (via the bundled claude CLI, no per-use cost), so no Anthropic API key is required. Adding one is only an optional, faster path for the writing features. Nothing is shared with anyone.

First run: what to expect

From download to your first evaluated role. The installer and the in-app Launchpad do the heavy lifting; you mostly review and confirm.

  1. Install. Run the installer (a few clicks; it bundles Node, Chromium, Claude Code, and Git). If SmartScreen warns "unknown publisher," click More info -> Run anyway. Restart if it asks (that puts Git on your PATH).
  2. Launch. Open trajecktory from the desktop or Start Menu shortcut, or tell Claude in Code mode "Start the live dashboard." It opens at http://localhost:3333.
  3. Take any update. If an "Update available" banner appears, click Update now. It is one-click and updates system files only; your CV, profile, tracker, and reports are never touched.
  4. Work the Launchpad. The Setup tab guides you with a readiness meter: paste your CV (or a LinkedIn URL, or upload a file), then confirm your identity, target roles, your edge, comp, location rules, evaluation tuning, and companies to track.
  5. Sign in to Claude. Click "Sign in to Claude" in the sidebar once. This is what lets Evaluate and Scan run, on your own Claude plan. No Anthropic API key is required; adding one is only an optional, faster path for the writing features.
  6. Run your first search. From the sidebar: API Scan (free, no AI) pulls fresh roles from the job boards, then the liveness gate removes dead postings before evaluation. Review the queue, evaluate the strongest roles, and let trajecktory tailor a resume and cover letter, and track it. It schedules the follow-ups.

Fuller walkthrough: docs/onboarding/first-run.md. Illustrated guides: docs/onboarding.

The dashboard, tab by tab

  • Today a weekly time-blocked cadence, a focus timer, a streak, and your to-dos.
  • Overview your weekly scorecard: inputs (verified touches, connects, cadence, screens booked) against floors, outcomes (warm vs cold reply, expired-before-action) against sourced benchmarks.
  • Pipeline & Tracker browse, filter, and sort every application; a per-role drawer renders the full A-to-G evaluation as a cheat sheet.
  • Follow-Ups the stale-thread action queue, with coach intelligence on what is overdue and a drafted, editable nudge for each.
  • Contacts an in-network (target-talent) contact CRM plus your referral network, with AI-drafted, voice-matched, editable outreach.
  • Social a Visibility tracker (your LinkedIn Social Selling Index), a connection queue, an engagement drafter, and a Posts composer: write your own or have Claude draft them, edit either, queue them, and schedule to LinkedIn and X through Buffer.
  • Interview per-company prep and a live "click a cue" board for the round you are about to run.
  • Insights honest coaching analytics, plus an optional Gmail sync that logs replies and bounces.

How it works

  1. Onboard in the Launchpad: add your CV, profile, and target roles.
  2. Scan portals (or paste a single job URL); dead postings are liveness-gated out before any AI spend.
  3. Evaluate Claude reads each posting against your CV (reasoning about fit, not keyword matching) and writes a structured report. The headline score is derived by code from keyed dimensions, with pay rated but deliberately kept from moving the score.
  4. Tailor generate a per-role docx resume + cover letter, bullets and all.
  5. Track and act manage status, follow-ups, and in-network outreach from the dashboard; capture replies from Gmail.
  6. Learn honest insights show what is actually converting, so you target better over time.

Also runs in any agent CLI

Prefer the terminal? The same engine works headless. trajecktory follows the open agent skill standard, so it runs in Claude Code, Gemini CLI, or OpenCode: paste a job URL or use the slash commands. See docs/SETUP.md.

Tech Stack

  • Dashboard: Node/Express + React (esbuild), served locally on 127.0.0.1.
  • Agent: Claude Code (also Gemini CLI / OpenCode) with custom skills and modes.
  • CV: docx generation via adm-zip slot-swap and per-JD bullet tailoring, preserving your master template byte-for-byte outside the parts it tailors.
  • Scanner: Playwright + ATS APIs.
  • Data: local Markdown + YAML + TSV. No database, no cloud, no telemetry.

Origin

trajecktory began as career-ops by santifer (MIT), a CLI-first job-search tool he used to evaluate 740+ offers, generate 100+ tailored CVs, and land a Head of Applied AI role. trajecktory builds on that foundation and reshapes it into a dashboard-driven, full-pipeline command center.

Ethical Use

trajecktory is built for quality, not quantity. It never submits an application on your behalf: it fills forms, drafts answers, and generates resumes, then stops so you make the final call. It strongly discourages low-fit applications, because your time and the recruiter's are both worth respecting. A well-targeted application to five companies beats a generic blast to fifty.

Disclaimer

trajecktory is a local, open-source tool, NOT a hosted service. By using this software, you acknowledge:

  1. You control your data. Your CV, contact info, and personal data stay on your machine and are sent directly to the AI provider you choose. We do not collect, store, or access any of it.
  2. You control the AI. The default prompts instruct the AI not to auto-submit applications, but models can behave unpredictably. If you modify the prompts or use different models, you do so at your own risk. Always review AI-generated content before submitting.
  3. You comply with third-party ToS. Use this tool in accordance with the Terms of Service of the career portals you interact with. Do not use it to spam employers or overwhelm ATS systems.
  4. No guarantees. Evaluations are recommendations, not truth. AI models may hallucinate. The authors are not liable for employment outcomes, rejected applications, account restrictions, or any other consequences.

See LEGAL_DISCLAIMER.md for full details. Provided under the MIT License "as is", without warranty of any kind.

License

The code is licensed under MIT. "trajecktory" is the project's brand name: forks are welcome under MIT, but please use your own product name and do not imply endorsement.

↑↓ navigate openesc close