What problem does it solve?
When an agent discovers a bug in the Lingtai project, it needs a safe, consistent way to file a GitHub issue without overstepping human authority or leaking credentials. This reference defines the decision flow for turning a drafted issue report into an actual filed issue.
Core Features & Use Cases
- Mandatory human consent boundary: The agent never files a GitHub issue without an explicit per-issue yes from the human, and drops the matter if declined.
- Read-only gh probe: Detects whether the GitHub CLI is installed and authenticated (via existing gh auth or a session-provided GH_TOKEN) before proposing a filing path.
- Two filing paths: Path A files directly with gh issue create using a body tempfile, while Path B hands the human a paste-ready title and body for manual filing.
- Token hygiene and proactive surfacing: Keeps tokens out of logs and files, and prompts the agent to surface bugs it discovers without waiting to be asked.
- Use Case: An agent spends several turns diagnosing an undocumented workaround, drafts a structured report, probes gh auth, asks the human for permission, and files the issue to Lingtai-AI/lingtai, quoting back the issue URL.
Quick Start
Ask the agent to check whether gh is authenticated and, with your approval, file the drafted issue report to the Lingtai-AI/lingtai tracker.
