Fix prompts
agentlint hands you the remediation, not just the diagnosis — and never calls an AI itself.
The loop
Each of the 30 rules ships a static prompt template, parametrized with your detected package manager, language, and workspace layout. Print one consolidated, copy-pasteable prompt for everything that failed:
agentlint prompt | pbcopy
Paste it into Claude Code, Cursor, Copilot, Codex — any agent. The prompt orders findings by score impact, includes anti-gaming ground rules (fix the cause, don't fake the signal), and ends with a verification step that re-runs the audit.
Narrowing to specific rules
agentlint prompt --rule agents-md-exists agentlint prompt --rule codeowners --rule pr-template
In the dashboard
Pushed runs keep their prompts. Open any run page and copy the consolidated prompt — or a single rule's prompt — straight from the browser. Useful when CI catches a regression and the fix should happen on a teammate's machine.
No AI inside
The prompts are generated from static templates — never by an LLM. agentlint makes zero AI calls, collects no telemetry, and works offline. The AI is yours; agentlint just briefs it.