BashBard: The Agentic Shell Guard

BashBard is an AI-powered Linux assistant that turns plain English into safe shell commands with danger checks and approval gates, reducing downtime and boosting reliability.

Comment

Inspiration

Downtime is one of the most expensive problems in IT. A single wrong Linux command can wipe servers, crash production, and cost a business millions. In Release 1, BashBard started as a tutor—helping beginners by fixing typos and explaining commands in plain English. That was good for learning, but real IT environments demand much more: speed, safety, and resilience. This challenge inspired Release 2: BashBard – The Agentic Shell Guard, built not just to teach, but to protect businesses from catastrophic IT mistakes.

What it does

BashBard is a safety-first AI assistant that combines natural language with strict risk control.

  • Understands plain English like: “show hidden files in /etc” → outputs correct command.
  • Fixes broken commands by analyzing error messages and replanning.
  • Every command is scanned by a danger-check engine to detect destructive patterns.
  • Risky actions trigger a human approval gate with safer alternatives.
  • Runs in dry-run mode by default, so nothing harmful executes without explicit approval.

Result: IT teams solve problems faster, avoid disasters, and deliver more reliable service to clients.

How it helps businesses

  • Faster response times: tickets close in minutes instead of hours.
  • Lower downtime: fewer outages, higher uptime SLAs.
  • Risk prevention: catastrophic errors stopped before execution.
  • Knowledge transfer: junior engineers get explanations with every fix.
  • Direct business impact: less downtime → lower cost → happier customers.

How we built it

  • Release 1: Python CLI + simple LLM prompts for translation and learning.

  • Release 2:

    • Migrated to LangChain + LangGraph for agentic workflows.
    • Built a safety module to block destructive commands.
    • Added multi-provider LLM support (OpenAI GPT + Google Gemini).
    • Developed an interactive CLI with argparse/click for approvals and dry-runs.

Challenges we ran into

  • Balancing usability vs. safety: keeping it easy to use while enforcing strict guardrails.
  • Building an approval workflow that is interactive but efficient.
  • Supporting multiple AI providers with a unified experience.

Accomplishments that we're proud of

  • Evolved BashBard from a teaching tool into a business-grade safety layer for IT operations.
  • Built a live, working CLI that already prevents destructive mistakes in real time.
  • Delivered Release 2 exclusively for IT HACKS 25, proving speed, focus, and ability to ship.

What we learned

  • IT support tools gain trust only when humans remain in control.
  • Blending AI automation + explainability makes the solution valuable to both experts and beginners.
  • Businesses care about risk prevention as much as efficiency.

What's next for BashBard: The Agentic Shell Guard

  • Slack and Jira integrations so teams can call BashBard directly inside their workflows.
  • Docker sandbox execution to test commands safely before touching production.
  • Enterprise dashboard with audit logs, compliance reporting, and risk analytics.
  • Expanding beyond Linux: future versions will support Windows CMD and PowerShell, unlocking another massive share of IT support environments worldwide.

Built With

Share this project:

Updates