How Career‑Ops Integrates with Different AI Coding CLIs: A Developer Guide

Career‑Ops integrates with AI coding CLIs through a thin wrapper architecture where AGENTS.md provides a single canonical skill definition and CLI‑specific markdown files translate native commands into uniform career‑ops invocations.

The santifer/career-ops repository demonstrates a clean, AI‑agnostic approach to CLI integration. Rather than baking vendor logic into the core, the project isolates business logic and exposes it through lightweight adapters. This pattern lets developers use the same job‑tracking and application workflow across Claude, Codex, OpenCode, Grok, Qwen, Kimi, and other agent‑style tools.

---

How Career‑Ops CLI Integration Works: Core Architecture

The Career‑Ops AI CLI integration follows a strict separation of concerns. Understanding this architecture helps you extend support to new agents or debug invocation issues.

AGENTS.md: The Central Skill Contract

Every supported CLI ultimately points to AGENTS.md in the repository root. This file defines:

• Skill name: career‑ops
• Prompt template: The workflow driver
• Sub‑modes: scan, apply, interview‑prep, pipeline, add, set‑status, etc.

No CLI‑specific code lives here. It is pure, portable skill definition.

CLI‑Specific Wrappers: Minimal Glue Code

Each supported CLI receives a short markdown file that imports AGENTS.md and adds invocation syntax:

• CLAUDE.md – Claude Code interactive and headless patterns
• CODEX.md – OpenAI Codex wrapper with exec command mapping
• OPENCODE.md – OpenCode slash‑command and run syntax
• KIMI.md – Kimi CLI integration
• GEMINI.md – Legacy Gemini wrapper (redirects to Antigravity CLI)

These files typically contain 10–30 lines: import statement, command syntax, and optional flag translation notes.

---

Running Career‑Ops Across Different AI Coding CLIs

The Career‑Ops commands remain identical regardless of which CLI you use. The wrappers only translate how you enter the skill.

Claude Code Integration

Interactive mode – start the CLI, then invoke directly:

claude

# then type:

/career-ops scan

Headless / batch mode – pass via -p flag:

claude -p "career-ops scan --query senior backend"

The CLAUDE.md wrapper documents both patterns and notes that slash commands require the leading /.

Codex CLI Integration

Interactive mode:

codex

# then type:

career-ops scan

Headless execution via exec:

codex exec "career-ops scan --data engineer"

Notice: Codex omits the leading slash in interactive mode, and headless uses exec instead of -p.

OpenCode CLI Integration

Interactive mode uses a leading slash:

opencode

# then type:

/career-ops apply --url https://example.com/job

Headless mode uses run:

opencode run "career-ops apply --url https://example.com/job"

The OPENCODE.md wrapper explicitly maps these two invocation styles to the same underlying skill.

Antigravity, Grok, Qwen, and Other CLIs

These tools share a common pattern: -p flag for headless prompt passing.

Antigravity CLI:

agy -p "career-ops pipeline"

Grok Build CLI:

grok -p "career-ops scan"

Qwen:

qwen -p "career-ops add --role 'Senior ML Engineer'"

GitHub Copilot CLI:

copilot -p "career-ops interview-prep --company Acme --role Engineer"

The GEMINI.md wrapper redirects Gemini users to the Antigravity CLI command, maintaining backward compatibility without duplicating logic.

---

Token‑Free and Headless Modes

Some Career‑Ops operations require zero LLM tokens. The CLI wrappers expose these directly by calling Node.js scripts:

Mode	Script	CLI Example
ATS‑free scanner	node scan.mjs	Any wrapper can invoke
Status tracker	node set-status.mjs	Direct script call

This design means the AI coding CLI is optional for deterministic operations. You can run node scan.mjs standalone, or wrap it for consistency within your preferred agent workflow.

---

Extending Career‑Ops to New AI Coding CLIs

Adding support for an emerging CLI requires only three steps:

1. Create [CLI_NAME].md in the repository root.
2. Import AGENTS.md at the top of the file.
3. Document the invocation syntax – interactive prefix (if any), headless flag, and any quirks.

The core logic in career‑ops remains untouched. Data contracts, trackers, and pipeline stages need no modification.

Reference docs/SUPPORTED_CLIS.md for the canonical table of all supported CLIs and their current invocation patterns.

---

Summary

• AGENTS.md provides the single source of truth for all Career‑Ops AI CLI integration.
• Thin wrapper files (CLAUDE.md, CODEX.md, OPENCODE.md, etc.) translate CLI‑native syntax into uniform skill calls.
• All CLIs expose identical commands: scan, apply, pipeline, interview‑prep, add, set‑status.
• Token‑free modes bypass the LLM entirely by invoking Node.js scripts directly.
• New CLI support requires ~10 lines of markdown documentation, zero core code changes.

---

Frequently Asked Questions

Does Career‑Ops require a specific AI coding CLI?

No. According to the santifer/career‑ops source code, the repository is explicitly AI‑agnostic. You can use Claude, Codex, OpenCode, Antigravity, Grok, Qwen, Kimi, or any future CLI that supports skill definitions. The core logic lives in AGENTS.md; wrappers only handle translation.

Why do some CLIs use /career-ops and others use career-ops?

The leading slash is a CLI‑specific convention. Claude Code and OpenCode use /career-ops for interactive slash commands, while Codex omits it. The wrappers in CLAUDE.md and CODEX.md document these differences so users know the correct syntax for their tool.

Can I run Career‑Ops without any AI CLI installed?

Yes. Operations like scan and set-status are token‑free and execute via node scan.mjs or node set-status.mjs. The AI CLI wrappers are convenience layers; the underlying scripts run independently with Node.js.

How do I add support for a new AI coding CLI?

Create a new markdown file in the repository root (e.g., NEWCLI.md), import AGENTS.md, and document the interactive and headless invocation patterns. See OPENCODE.md or KIMI.md as minimal templates. No changes to skill logic or data contracts are required.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}