First public release. Three agent skills on one gateway-agnostic engine — audit, clean, and harden a WordPress site end to end with an AI agent (Claude Code, Cursor, Codex, or any MCP client), over SSH + WP-CLI or any WordPress MCP (SproutOS recommended). Free & open source (AGPL-3.0).
Install
git clone https://github.com/adityaarsharma/wordpress-malware-removal.git
cd wordpress-malware-removal
./install.sh # installs all three skills into ~/.claude/skills--dir PATH for Codex/Cursor/other · --skill <name> for one · --uninstall. Helper scripts need Python 3.8+, zero third-party dependencies.
The three skills
- wp-security-audit — read-only "how hackable am I?" A–F posture score across access hygiene, known vulns (free Wordfence feed), config/exposure, code risk (SAST), and update status, with a ranked, honest fix list. Touches nothing.
- wp-malware-removal — 12-layer forensic scan → no-false-positive verify → quarantine-first removal → prove clean → names the entry vector. Catches the stealth infections in-WordPress scanners miss: cloaking, web-root shadow folders, DB-resident payloads, single-file backdoors, self-healing cron reinjection.
- wp-hardening — 20 idempotent, reversible hardening actions classified SAFE / RISKY / MANUAL; sentinel-wrapped server-config edits, per-action verify + rollback. Auto-applies safe fixes, asks before risky ones.
Safety DNA
Backup is the user's and is enforced (never self-served) · quarantine before delete · health-checked after every destructive step with auto-rollback on regression · unsure → NEEDS_HUMAN, never auto-delete · honest coverage gaps over a silent "all clear."
Detection, hardening, and audit methodology is sourced — see docs/research-2026-07-wp-security.md.