Security audit
Security checks for vulnerabilities and agentic risk
This skill is a disclosed SQL diff review guide that reads local SQL changes and reports issues without editing unless the user explicitly approves fixes.
Install only if you want the agent to inspect local SQL diffs and quote relevant SQL lines in its findings. It should not modify SQL during review unless you explicitly approve fixes afterward.
The description says to use the skill whenever the user asks about SQL antipatterns in a diff, even if they do not name the skill. That activation condition is broad enough to overlap with ordinary conversational requests and could cause unintended invocation without a clear boundary or negative examples in the manifest description.
Several listed triggers are short, natural phrases a user might say casually, which increases the chance of accidental activation. While the skill is SQL-focused, the trigger section does not fully constrain scope with explicit non-trigger examples beyond mixed-diff review.
Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.
- ❌ Trying to parse SQL with a full grammar — use ripgrep on the diff text. The skill is heuristic by design.
- ❌ Flagging every existing antipattern in a modified file. Stay scoped to the diff for modified files; full-file scan only for new files.
- ❌ Editing SQL during the review pass. Report first, **always**.
- ❌ Suggesting autofixes for `BLOCKER` items without asking. SQL deploys to production; user confirmation gates the change.
- ❌ Listing a `nit` when there's a `BLOCKER` next to it. Lead with severity.
- ❌ Hand-wavy findings without `file:line`.
- ❌ Reporting a raw ripgrep hit without reading the surrounding procedure. A pattern match is a candidate; the proc's full context sets the severity (see [Pattern hit ≠ finding](#pattern-hit--finding)).