Agent actions
It runs commands and shows real results in the chat.
Many developers search for a sql injection test online when they want to verify their own code, but CodeSpar is not an online service. It runs tools on your PC against your own staging copy and shows the results in the chat. Similarly, people often look for an online sql injection scanner, yet CodeSpar is not an online scanner. It handles the workflow using your local shell commands on a staging copy, ensuring you test only what you own.
It runs commands and shows real results in the chat.
It edits files and shows diffs before you accept changes.
You choose the mode; in Ask first you approve changes.
The agent finds database calls and rewrites queries using parameterised statements. It checks output encoding and verifies security headers on your staging site.
grep -r "SELECT" --include="*.py3 matchesinjection-test-plan.txtread 12 linessrc/query.pysavedrun testsexit 0http_get https://staging.codesmite.net200 okThe agent uses grep to search the source code for database calls and edit_file rewrites the code to use parameterised queries while showing you the exact diff. This is a test. You review the changes in Ask first mode before accepting the update. Shell runs your unit tests to confirm the fix works without breaking the app. Set up a staging copy with fake data and a backup for safe testing. Give the database account low rights to prevent accidental damage during the sql injection test online. Tracing dom xss from source to sink requires careful analysis of the application logic. Use the sql injection scanner to verify your findings and check the results.
The agent uses http_get to fetch a page from your staging site and reads the status code, so it verifies the Content-Security-Policy header is present on the page and confirms security posture. You review the plan. Plan only writes a detailed plan and runs nothing at all during this phase. You switch to Ask first mode to apply fixes safely and correctly. You keep the work to your own staging site for privacy. The agent reads the file to confirm the output is properly encoded. The editor notes the changes in the diff for your review. The test database account has the lowest rights that work effectively, and you might check the cvss score for sql injection to prioritize findings.
It is not a hosted service or plugin. It is meant only for your own files, PC and servers you are authorised to use. You keep the work within them. The agent does not enforce scope. Shell, web and SSH are not folder-bound. There is no guarantee that code, scans or analysis are correct.