Security plugins

Skill + SQL checker + a mod that refuses GRANT ALL or write grants to anon or PUBLIC, and role grants to anon/authenticated. /grants-check shows how to run the checker.

1 skill

Route any artifact, investigate it feature-first, keep every claim tied to evidence, and rebuild observed behavior cleanly.

1 skill

Wires the hosted Agentic Trust gateway MCP so any Claude Code agent can route privileged calls through the open, anonymous, rate-limited enforcement layer that verifies MM-issued CIS-4/CIS-8004 credentials on Concordium. No key, no auth, no account.

1 MCP server

Defensive security for everyday shipping: 10-minute threat models, per-entry-point review checklist (injection, authZ/IDOR, secrets, crypto, uploads), dependency/supply-chain routine, incident basics, and deep-review orchestration for full audits. Authorized-defensive scope only.

1 skill

Mechanical safety guards for agents: hooks that refuse a dangerous command and tell the agent what to do instead. Guards: no inline node/bun eval (-e, -p), which turns any library that forks with execArgv into a fork bomb; no package install inside a git submodule whose packages are workspaces of the superproject, which shadows its dependencies; no command that prints a secret value into the transcript (an env dump under secrets run, a keychain read, the store token, bws secret get)

hooks

A PostToolUse hook that masks .env values, token shapes and secret-named values in Read, Bash and Grep results before the model, the transcript and the terminal see them.

hooks

Security & abuse review skill: economic abuse, races, auth-flow abuse, deployment boundaries, dependency triage.

1 skill

Audit WebRTC and DNS leaks on macOS behind Clash Verge / TUN, trace where system DNS really comes from, and isolate the foreign browser without breaking domestic WebRTC.

1 skill

Read-only security audit of your own Linux VPS over SSH (Lynis, docker-bench-security, Trivy, nmap) with a prioritized report and optional, approved fixes with rollback.

1 skill

Cloudflare's open-source security-audit skill for application source code (MIT, by Cloudflare). Pinned to a reviewed commit.

1 skill

Finds and removes invisible or deceptive Unicode (zero-width, bidi/Trojan Source, tag-character and variation-selector payloads, blank-glyph identifiers) while keeping emoji and complex scripts intact.

1 skill

Confidentiality toolkit: smart routing across cryptoid (Hugo encryption), pagevault (HTML/file encryption), and gpg (file encryption/signing)

1 skill

Anthropic: Sicherheits-Hooks: Regex-Warnungen bei Edits, LLM-Review des Diffs am Rundenende und bei git commit/push (verbraucht API-Kontingent).

hooks

Security reminder hook that warns about potential security issues when editing files, including command injection, XSS, and unsafe code patterns

hooks

Security reminder hook that warns about potential security issues when editing files, including command injection, XSS, and unsafe code patterns

hooks

Security review for Claude-generated code. Pattern-based warnings on edits, LLM-powered diff review on Stop, and an agentic commit reviewer that catches injection, XSS, SSRF, hardcoded secrets, and 25+ other vulnerability classes.

hooks

Essential Auth0 skills including quickstarts, migration from other providers, and Multi-Factor Authentication (MFA).

1 skill

Google Cloud Armor WAF log patrol for Claude Code, straight from Cloud Logging. Summarize enforced DENYs per rule, flag home-region false positives, review preview (dry-run) rules, and get a one-shot daily_brief health check. All tools are read-only. Requires CLOUDARMOR_PROJECT and a GOOGLE_APPLICATION_CREDENTIALS service-account key file on the local machine.

1 MCP server

Block Bash commands that find processes by name or command-line text and kill them (pkill, killall, xargs kill, kill $(pgrep …)), and tell Claude to stop processes by PID instead.

hooks

Author Lacework FortiCNAPP LQL queries and custom policies from the official LQL Reference Guide.

1 skill

MRVA end to end: variant query authoring, target lists, controller repository, gh-mrva / VS Code / REST runs, SARIF triage, disclosure.

1 skill

Block git force pushes (--force, -f, --force-with-lease) before the agent runs them, and tell it to ask the user first.

hooks

Governs the tools Claude Code runs on a developer machine through your organization's Meilynx proxy, and seals every decision into your audit chain.

hooks

Repeatable consent-scenario website scan, sourced GDPR/ePrivacy/PECR rule catalogue, data usage map, and fixes on your go-ahead.

1 skill

Blocks pgrep/pkill commands that match the agent's own process or poll in a loop.

hooks

Guard rails for destructive operations. plan measures first, apply quarantines with a sha256 manifest, restore verifies, purge is the only irreversible command.

1 skill

Cloudflare's security-audit skill, installed from cloudflare/security-audit-skill at c1c8a8c (MIT). Nothing is redistributed here.

1 skill

Manage secrets via the 1Password CLI (op): read, create, and inject credentials through op:// references so other skills get least-privilege access without exposing values.

1 skill

Pre-submit scan, audit, validate, and package AgentSkills using the local codeql-sdk/clawhub-audit scanner and OpenClaw skill validation, emitting JSON/SARIF reports.

1 skill

Apply OpenClaw's MITRE ATLAS-mapped threat model to evaluate prompt injection, malicious skills, token theft, tool over-permission, and safe rollout of risky automations.

1 skill