Career-Ops

A local-first, multi-agent-CLI career operating system: discover and filter roles, evaluate fit, draft tailored artifacts, maintain pipeline state, prepare for interviews, and keep the human in control.

Career-Ops began as a job-search project built with Claude Code, but the current source is a portable skill router for Claude Code, Codex, OpenCode, Antigravity, Grok, Qwen, Kimi, and other Agent Skills-compatible CLIs. Kevin's integration is deliberately compact: Career Pipeline Workflow owns recurring execution, the career-ops skill owns invocation, and this page owns capability, version, and evidence boundaries. Source: santifer/career-ops README, supported-CLI guide, and skill at 92dea48b, 2026-08-11

Current Surface

Family Current capability
Discovery Public job-board/API scanning, custom portals, liveness checks, deduplication, and cheap title/location triage
Evaluation Structured role-fit reports, posting-legitimacy checks, comparisons, company research, and recommendation thresholds
Artifacts Tailored ATS CV/PDF/LaTeX, cover letter, email, outreach, and application-field drafts
Pipeline URL inbox, tracker, statuses, follow-ups, reply classification, outcomes, and a Go terminal dashboard
Interviews Company preparation, plan, practice, debrief, story-bank reuse, and red-flag analysis
Strategy Pattern analysis, adjacent titles, skill gaps, training/projects, contacts, and offer preparation
Runtime Natural-language or named-mode routing across major agent CLIs, opt-in plugins, tests/evals, and guarded system updates

The upstream mode catalog is implementation detail behind the router; it does not require a wiki page per mode. The local skill selects and loads only the relevant context.

Operating Graph

Proof And Claim Boundaries

The reviewed 22.7-second, 1664×1080 source video on X shows a real stateful terminal workbench: 516 offers, scored/status/salary views, a detailed Datadog role brief, CV-match analysis, LinkedIn-change suggestions, and interview red-flag guidance. It proves the operator surface and output families. It does not prove that Career-Ops caused a hire, that its fit scores are calibrated for Kevin, or that every portal and automated field path remains live. Source: complete local video capture, inspected 2026-08-11

The viral X post reports that the system evaluated 700+ applications and helped its creator get a job; the current README reports 740+ listings, 100+ personalized CVs, and one role landed. These are author/reposter outcome claims, not controlled causal evidence. Kevin's route therefore requires current source checks, representative-role evaluation, and human review. Source: X/@Hesamation, 2026-04-05; upstream README, 2026-08-11

Data And Authority Model

Boundary Local rule
User layer Kevin's CV, profile, facts, custom rules, story bank, trackers, contacts, offers, outcomes, and reports remain user-owned and are never overwritten by system updates.
System layer Procedures, templates, scripts, providers, skill projections, and tests may be refreshed only with a diff, validation, and rollback path.
Local-first Career files live locally, but context selected for an agent run is sent to Kevin's chosen AI provider. Upstream says it collects no project telemetry.
Applications The system may evaluate and draft. Kevin performs the final send, submit, or click.
Resume Generated variants are proposals. raw/career/resume-kevin-liu.tex changes only after Kevin approves a factual diff.
Sources Use public permitted endpoints, preserve receipts, respect terms, and never turn scanning into spam or prohibited bulk automation.

Portal Counts Are Mutable

The current README describes 100+ companies and 45+ search queries; the current hosted site says 150+ company portals. These are different mutable surfaces and are preserved as such—not collapsed into one timeless support count. Exact provider support comes from the pinned providers/ tree and docs/SUPPORTED_JOB_BOARDS.md, while a live run still has to record liveness failures. Source: upstream README and hosted site, captured 2026-08-11

Invocation

For a new standalone upstream install, the current entry point is:

npx @santifer/career-ops init

Kevin-Wiki already carries a local career-ops skill, so agents should invoke that skill and Career Pipeline Workflow rather than installing blindly. In Codex, use a natural-language mode request or codex exec; native slash-command availability is not assumed. Source: upstream README and docs/CODEX.md, 2026-08-11

Version Snapshot

Surface Captured authority
Release career-ops-v1.26.0, published 2026-08-10
Source HEAD 92dea48b209a9791b884e2a19ee111d0e2f098f0
GitHub signal 63,460 stars at enrichment time; mutable popularity evidence, not quality proof
License MIT
Runtime Node scripts, Playwright, Go/Bubble Tea dashboard, Markdown/YAML/TSV/SQLite-derived state
Evidence pack .brain/artifacts/x/2040820834428658008/career-ops/

Kevin-Wiki Integration

  1. career-ops is the executable router; Career Pipeline Workflow is the recurring orchestration owner.
  2. Career Profile supplies live role, company, location, and compensation constraints.
  3. Resume, Kevin B. Liu and raw/career/resume-kevin-liu.tex supply verified candidate facts; variants never silently rewrite the master.
  4. Recruiter Discovery Workflow handles relationship discovery; Career Ops handles role and application state.
  5. Every run leaves sources, scoring rationale, proposed artifacts, state transitions, failures, and human approvals in a receipt.

Timeline