When scanning the code for security vulnerabilities using the Aikido MCP server:

Identify all files that were generated, added, or modified in this session (or that the user has mentioned).
Prefer aikido_scan_paths whenever the files exist on disk (files you just wrote/edited, files reported by git status/git diff, files the user points at). Pass the file paths (absolute, or relative to a root you provide) — do not read the files first, the server reads them itself, so passing content would only waste context. Only fall back to aikido_full_scan for code that is not saved to disk yet (a proposed diff, a pasted snippet, freshly generated code you haven't written out) — for that tool, read each file's full content and pass it along.
Stay within the 50-file limit per request for either tool — batch into multiple calls if needed.
If any security issues are found:
- Explain each issue clearly: title, description, severity, file location, and line numbers.
- Apply fixes guided by the remediation provided by Aikido.
- After applying all fixes, re-run the same tool (aikido_scan_paths for on-disk files, aikido_full_scan otherwise) to verify that the issues were resolved and no new issues were introduced.
- Stopping the loop: If you can explain why the applied fix is safe (e.g. the fix correctly addresses the finding and the remaining scan output is a false positive or acceptable), you may stop and report to the user. Otherwise, repeat the fix-and-rescan cycle up to 3 attempts; if issues remain after that, report them to the user instead of continuing.
Report the final scan result to the user — confirm all clear or list any unresolved issues with explanation.

If the Aikido MCP server is not available or fails to start, inform the user:

The Aikido MCP server is required for security scanning but is not available. Run /aikido:setup to configure it.