Skip to content

helmcode/agentcrew-job-search

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1 Commit
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

🎯 AgentCrew Job Search

An autonomous AI agent team that searches job portals daily, scores offers against your profile, and delivers a ranked email summary every morning.

Built on AgentCrew — open-source multi-agent AI orchestration platform.


How it works

Every day at 08:00 (configurable), four agents collaborate autonomously:

[Schedule 08:00]
      ↓
[coordinator]  ← leader, orchestrates the workflow
      ↓
[scraper]      ← browses InfoJobs, Indeed, LinkedIn via web search
      ↓
[analyst]      ← scores each offer against your profile (threshold: 7/13)
      ↓
[reporter]     ← composes and sends the daily email summary

Each offer in the email includes:

  • 🏢 Title, company, location and modality
  • 🎯 Score breakdown (X/13)
  • 📝 2-3 line summary and key match reason
  • ⚠️ Alerts (e.g. requires English C1, 100% on-site)
  • 🔗 Direct link to the offer

Scoring system

Signal Points
Team management (direct or indirect) +3
KPIs / metrics / reporting +3
Operations / process improvement +3
Cross-functional coordination / stakeholders +2
Compatible sector (BPO, telco, contact center, utilities) +2
Pure sales role (no management) -3
Pure technical role (no management) -3
Junior / Intern / Entry-level -5
CEO / C-Level / Director General -5

Minimum threshold to appear in the report: 7/13


Prerequisites

  • AgentCrew installed and running
  • Docker
  • A Gmail account with App Password enabled (or any SMTP server)

Setup

1. Install AgentCrew

curl -fsSL https://agentcrew.helmcode.com/install.sh | bash

2. Clone this repo into your AgentCrew workspace

git clone https://github.com/helmcode/agentcrew-job-search \
  ~/.agentcrew/teams/job-search
cd ~/.agentcrew/teams/job-search

3. Build the custom Docker image

The scraper agent needs Playwright MCP. Build the custom image:

docker build -t helmcode/job-search-scraper:latest .

4. Set up a local registry

AgentCrew pulls images from a registry. Run a local one:

# Start the registry
docker run -d -p 5000:5000 --name registry --restart always registry:2

# Tag and push
docker tag helmcode/job-search-scraper:latest localhost:5000/job-search-scraper:latest
docker push localhost:5000/job-search-scraper:latest

# Also tag with the Docker bridge gateway IP (used by AgentCrew containers)
# Check your gateway: docker network inspect agentcrew_default | grep Gateway
docker tag helmcode/job-search-scraper:latest 172.20.0.1:5000/job-search-scraper:latest
docker push 172.20.0.1:5000/job-search-scraper:latest

Configure Docker to trust the local registry (no HTTPS):

echo '{"insecure-registries": ["172.20.0.1:5000"]}' | sudo tee /etc/docker/daemon.json
sudo systemctl restart docker

⚠️ Replace 172.20.0.1 with your actual Docker bridge gateway IP.

5. Configure email credentials

cp .env.example .env

Edit .env with your SMTP credentials. For Gmail, generate an App Password.

6. Create the team in AgentCrew

Open your AgentCrew UI and create a new team:

Field Value
Provider Claude Code
Team Name job-search
Workspace Path /path/to/.agentcrew/teams/job-search
Custom Agent Image 172.20.0.1:5000/job-search-scraper:latest

7. Create the agents

In AgentCrew Settings → Sub-Agents, create the three workers using the files in agents/:

  • agents/coordinator.md → name: coordinator (leader)
  • agents/scraper.md → name: scraper
  • agents/analyst.md → name: analyst
  • agents/reporter.md → name: reporter

8. Create the schedule

In AgentCrew → Schedules → New Schedule:

Field Value
Name job-search-daily
Team job-search
Frequency Daily at 08:00
Timezone Your timezone
Prompt Start the daily job search workflow. Search for job offers, score them and send the email report.

Customising for your profile

Edit agents/coordinator.md to change the search queries and location. Edit agents/analyst.md to adjust the scoring criteria and candidate profile. Edit agents/reporter.md to change the email format.


Known issues & fixes

Claude Code 2.1.92+ silent failure

If your agent starts, runs for ~30 seconds, exits with code 0 but produces no output, you are hitting a known issue:

  • Root cause: AgentCrew API launches agent containers as User: 0:0 (root). Claude Code 2.1.92+ blocks --dangerously-skip-permissions when running as root. Even after gosu drops to a non-root uid, HOME still points to /root which is not writable by the agent user. Claude silently exits with code 0 when it cannot write ~/.claude/.
  • Fix in this Dockerfile: We pin Claude Code to 2.1.90, set chmod 777 /home/agentcrew, and inject HOME=/home/agentcrew via the sidecar.
  • Upstream fix: A PR is open on agent_crew_api to stop setting User: "0:0" when launching agent containers.

Adding more job portals

Edit agents/scraper.md and add any portal to the search list. The rest of the pipeline does not need changes.


License

MIT

About

Autonomous job search agent team built on AgentCrew

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors