Agent handoff.
aislop fix resolves the safe stuff. For the rest (as any casts, oversized functions, complex refactors) it hands off to your coding agent with a prompt containing every remaining diagnostic and where to find it.
How it works
Run aislop fix first. It auto-resolves everything that's mechanically safe. Whatever's left is where judgment starts.
Then run aislop fix --<agent>. aislop builds a prompt listing every remaining diagnostic with file paths, line numbers, rule IDs, and severity, then launches your agent's CLI with that prompt pre-filled.
npx aislop fix npx aislop fix --claude Skip handoff entirely: install a hook
aislop hook install wires aislop into the agent's own lifecycle. Every time the agent edits a file, aislop scans the changed files and feeds back score + findings + next steps. The agent self-corrects before the turn ends. No fix call needed.
npx aislop hook install --claude Supported agents
Fourteen agents supported. The flag invokes the agent's local CLI, so you need that agent installed.
Agent-agnostic: --prompt
If your agent isn't listed, or you want to paste the prompt into a web UI, use --prompt. aislop prints the full handoff prompt to stdout; pipe it wherever:
npx aislop fix --prompt | pbcopy npx aislop fix --prompt > prompt.txt What's in the prompt
- The current score and label (Healthy / Needs Work / Critical)
- Every remaining diagnostic, grouped by file, with line numbers and rule IDs
- Short description of what each rule catches, so the agent knows why the issue matters
- Priority hint: errors first, then high-severity warnings, then the rest
- Instruction to leave auto-fixable items alone (aislop fix handles those)
Install aislop as an agent skill
Run npx skills add scanaislop/skills and every coding agent on your machine (Claude Code, Cursor, Codex, Gemini CLI, Windsurf, Cline, and 30+ more) picks up the quality gate every turn, not just at handoff. The agent runs the scan, auto-fixes the mechanical findings, and only reports “done” when the bar is met.