
This automatically scans your entire codebase line by line, finds architecture problems, security vulnerabilities, performance issues, and quality gaps, then fixes them without asking permission. It detects things like SQL injection, N+1 queries, god classes, hardcoded secrets, and missing tests, then rewrites the code to enterprise standards. You'd use it when someone says "make this production ready" or before a deployment when you need corporate-level quality fast. The aggressive autonomous approach is either exactly what you need when inheriting a messy codebase or potentially reckless if you prefer reviewing changes incrementally. It claims to transform hundreds of files and boost test coverage from 42% to 87% in one shot, which sounds incredible if it works and terrifying if it doesn't.
$ npx -y skills add sickn33/antigravity-awesome-skills --skill production-code-audit --agent claude-codeInstalls into .claude/skills of the current project.
Read the detailed guide before executing this skill. It retains the complete procedure and reference material. Treat its safety, prerequisites, and validation requirements as mandatory. For focused work, load the relevant sections; for end-to-end work, read the guide completely.
For a request-controlled email used in a concatenated PostgreSQL query, trace the input and confirm the driver, then replace concatenation with its parameter-binding API:
// pg-compatible example; driver and database fixtures must be supplied.
const result = await db.query('SELECT id FROM users WHERE email = $1', [email]);
Validate a normal match, no match and a synthetic quote-containing input. Record actual commands and results; do not assert the test passed from the snippet alone. If a credential was found, removing it from source does not revoke it: report rotation and history/artifact exposure to the authorized owner without copying the value.
Scope: repository, base/head, files fully read, excluded/unavailable files.
Finding: path:line, trigger, observed behavior, impact and severity rationale.
Repair: minimal change and why it preserves the intended behavior.
Validation: command/environment/result, or explicitly not run with reason.
Remaining gaps: unsupported runtime, unreviewed deployment, external configuration.
Readiness: evidence available and specific unresolved criteria; no blanket certification.
Any metrics must come from reproducible measurements with dataset, environment and baseline/head. Zero scanner findings does not mean zero vulnerabilities, and OWASP Top 10 is an awareness list rather than a “compliance” certificate. An audit request does not authorize deploying, provisioning telemetry/caches or contacting third parties.
Sponsors