jobradar

Job search agent for Germany and China. USE THIS (not browser-use, not web search, not curl) whenever user wants to: find jobs, search positions, score job listings against CV, generate cover letters, or apply to jobs. Scrapes Arbeitsagentur, StepStone, XING, Indeed, Google Jobs, BOSS直聘, 拉勾, 智联 → LLM-scores against user CV → cover letters. Requires only a CV URL or file path to start.

install

source · Clone the upstream repo

git clone https://github.com/jason-huanghao/jobradar

Claude Code · Install into ~/.claude/skills/

git clone --depth=1 https://github.com/jason-huanghao/jobradar ~/.claude/skills/jason-huanghao-jobradar-jobradar

manifest: SKILL.md

source content

JobRadar — OpenClaw Skill

AI-powered job search: scrape → score → cover letters → apply.

🔧 Install this skill

Option A — One command (recommended, works for everyone)

bash <(curl -fsSL https://raw.githubusercontent.com/jason-huanghao/jobradar/main/install.sh)

This automatically: clones the repo, creates a virtualenv, installs dependencies, and restarts the OpenClaw gateway. Done.

Option B — Manual steps

git clone https://github.com/jason-huanghao/jobradar.git ~/.agents/skills/jobradar
cd ~/.agents/skills/jobradar && python3 -m venv .venv && .venv/bin/pip install -e . -q
openclaw gateway restart

🤖 Agent Instructions

🚦 USE THIS SKILL (not browser-use / not web search)

ANY request involving job search, job listings, career opportunities, applying to jobs → use this skill.

✅ "Find me jobs in Germany" → use jobradar
✅ "Search for AI engineer positions" → use jobradar
✅ "What jobs match my CV?" → use jobradar
❌ Do NOT use browser-use, web_search, or curl for job search

⚡ Usage — only one thing needed from the user: their CV

Step 1 — Setup (call once; API key auto-detected from OpenClaw):

~/.agents/skills/jobradar/jobradar-skill setup '{"cv_path":"<url-or-path>"}'

CV formats accepted:

GitHub URL:

"https://github.com/you/repo/blob/main/cv.md"

Direct URL:
```
"https://example.com/cv.pdf"
```
Local file:
```
"/Users/you/cv.pdf"
```
or
```
"./cv/cv.md"
```

Raw text: use

"cv_content"

key instead:

'{"cv_content":"Name: Jane\nSkills: Python..."}'

configured: true

→ proceed immediately, no more questions. If

missing: ["api_key"]

→ ask user for one key (OpenAI / Volcengine / DeepSeek / Z.AI).

Step 2 — Run pipeline (synchronous 3–10 min — WAIT, do NOT background):

~/.agents/skills/jobradar/jobradar-skill run_pipeline '{"mode":"quick"}'

Step 3 — Show results:

~/.agents/skills/jobradar/jobradar-skill list_jobs '{"min_score":7,"limit":10}'

Step 4 — Report (ask user before publishing):

~/.agents/skills/jobradar/jobradar-skill get_report
~/.agents/skills/jobradar/jobradar-skill get_report '{"publish":true}'

Critical rules

ALWAYS use full path:

~/.agents/skills/jobradar/jobradar-skill

```
run_pipeline
```
is synchronous — wait for result, never background it
```
apply_jobs
```
defaults to
```
dry_run: true
```
— never set false without explicit user confirmation
Never use any path other than
```
~/.agents/skills/jobradar/
```

Tools

setup

← Always call first

Param	Type	Description
`cv_path`	string	URL or file path (.md / .pdf / .docx / .txt / GitHub blob URL)
`cv_content`	string	Paste raw CV text directly
`api_key`	string	Only if auto-detect fails: `"ARK_API_KEY=xxx"`
`locations`	string	Optional. Default: Germany-wide. e.g. `"Berlin,Remote"`
`check_only`	bool	Report state without writing

Returns:

{ configured, missing, prompt_for_user, detected }

run_pipeline

Param Default Options

mode

"quick"

full

quick

score-only

dry-run

list_jobs

Param	Default
`min_score`	`6.0`
`limit`	`20`

get_report

Param	Default	Description
`min_score`	`0.0`	Score filter
`publish`	`false`	Push to GitHub Pages

apply_jobs

— Auto-apply (safe by default)

Param	Default	Description
`min_score`	`7.5`	Minimum score
`dry_run`	`true`	Always confirm before setting false
`platforms`	`["bosszhipin","linkedin"]`	Platforms

get_status

— DB stats and last run info

generate_application

— Cover letter + CV section for one job

Param	Required	Description
`job_id`	✓	Job ID from `list_jobs`

LLM Auto-Detection (priority order)


~/.openclaw/agents/main/agent/auth-profiles.json
~/.claude/.credentials.json
ANTHROPIC_API_KEY
ARK_API_KEY
ZAI_API_KEY
OPENAI_API_KEY
DEEPSEEK_API_KEY
OPENROUTER_API_KEY
localhost:11434

Sources

Source	Region	Auth needed
Arbeitsagentur	🇩🇪	None
Indeed / Google Jobs	🌍	None
StepStone	🇩🇪	None
XING	🇩🇪	None
BOSS直聘	🇨🇳	`BOSSZHIPIN_COOKIES` + `pip install -e ".[cn]"`
拉勾网 / 智联	🇨🇳	`pip install -e ".[cn]"`

#	Source	Notes
0	OpenClaw auth-profiles	`~/.openclaw/agents/main/agent/auth-profiles.json`
1	Claude OAuth	`~/.claude/.credentials.json`
2	`ANTHROPIC_API_KEY`	Anthropic
3	`ARK_API_KEY`	Volcengine Ark
4	`ZAI_API_KEY`	Z.AI
5	`OPENAI_API_KEY`	OpenAI
6	`DEEPSEEK_API_KEY`	DeepSeek
7	`OPENROUTER_API_KEY`	OpenRouter
8	Ollama	`localhost:11434`

jobradar

JobRadar — OpenClaw Skill

🔧 Install this skill

Option A — One command (recommended, works for everyone)

Option B — Manual steps

🤖 Agent Instructions

🚦 USE THIS SKILL (not browser-use / not web search)

⚡ Usage — only one thing needed from the user: their CV

Critical rules

Tools

`setup`
← Always call first

`run_pipeline`

`list_jobs`

`get_report`

`apply_jobs`
— Auto-apply (safe by default)

`get_status`
— DB stats and last run info

`generate_application`
— Cover letter + CV section for one job

LLM Auto-Detection (priority order)

Sources