This is a submission for the GitHub Copilot CLI Challenge
What I Built 🚀
Introducing OpsPilot: an AI Incident Commander for the terminal, designed for real SRE incident workflows.
OpsPilot is a Node.js CLI that helps teams:
- analyze logs (
opspilot analyze) 🔎 - suggest next diagnostic commands with safety classification (
SAFE,CAUTION,DANGEROUS) (opspilot suggest) 🛡️ - generate incident timelines from shell history, git commits, and logs (
opspilot timeline) 🕒 - draft structured postmortems (
opspilot postmortem) 📝
The tool is recommendation-only and never auto-executes commands. It includes strict command safety checks and graceful fallback behavior when Copilot CLI is unavailable.
Demo 🎬
Suggested demo flow:
opspilot init --environment kubernetes --service payments-api --namespace prodopspilot analyze examples/sample-incident.logopspilot suggestopspilot timeline --log examples/sample-incident.logopspilot postmortem
Terminal Output Examples 💻
opspilot analyze examples/sample-incident.log
- Reading incident log...
✔ Log loaded.
- Asking GitHub Copilot CLI for incident analysis...
✔ Analysis received from Copilot CLI.
Incident Analysis
------------------------------------------------------------
## Incident Summary
...
## Likely Root Causes (ranked)
1. ...
2. ...
[OK] Saved analysis to /path/to/repo/.incident/analysis.md
opspilot suggest
- Reading shell history...
✔ Loaded 250 history entries.
- Asking GitHub Copilot CLI for command suggestions...
✔ Suggestions received from Copilot CLI.
Copilot Suggestions
------------------------------------------------------------
## Recommended Checks
- `kubectl get pods -n prod`
- `kubectl logs deploy/payments-api -n prod --tail=200`
Safety Classification
------------------------------------------------------------
[SAFE] kubectl get pods -n prod
[SAFE] kubectl logs deploy/payments-api -n prod --tail=200
[INFO] OpsPilot only recommends commands. It never executes them.
My Experience with GitHub Copilot CLI 🤖
I used GitHub Copilot CLI as the core reasoning engine for OpsPilot, not just a helper.
Key integrations:
runCopilotAsk(prompt) for deep incident analysis and postmortem drafting runCopilotSuggest(prompt) for next diagnostic command recommendations structured prompts that force useful Markdown output for incident response workflows
What stood out:
Copilot CLI made it much faster to turn raw logs/history into actionable incident guidance. Prompt quality mattered a lot; structured prompts produced consistently better, safer output.
Building robust wrapper logic around Copilot CLI (timeouts, command-format compatibility, auth failure handling) was essential for production-like reliability. Overall, Copilot CLI significantly reduced context-switching and accelerated my iteration loop while building a practical terminal-first SRE tool. 🎉
Top comments (0)