18,930 tagged Security, measured the same way as everything else here.
Browse within: cybersecurity 505bug-bounty 236agent 230generative-ai 178LangChain 177autonomous-pentesting 176hacking 176skills 131claude-ai 104LLM 103agentic-ai 87ai-security 76openclaw 73agents 64
Skill Claude CodeCodex
Inspect and fix email authentication (SPF, DKIM, DMARC, MTA-STS, BIMI) with Palisade. When to use: a domain's mail is being rejected or spoofed, a DMARC record needs writing or moving to p=reject, DMARC reports need reading, or an IT team or MSP wants ongoing monitoring across many domains. Not for sending or…
Skill Claude CodeCodex
Use when any task involves an API key, token, password, credential, cookie, private key, connection string, OAuth secret, or delegated access. Route credential storage and retrieval through AgentKeychain instead of chat, source files, logs, or repeated user prompts.
Skill Claude CodeCodex
Use when a request needs one of the evidence-labeled local Nymrel MCP utilities for bounded public web extraction/search, structured-data examples, command-pattern checks, local cryptographic receipt operations, quote formulas, or declarative UI JSON.
Skill Claude CodeCodex
Backend security guidelines dictating input validation, authentication enforcement, authorization, and database interaction safety.
Skill Claude Code
Read-only security review of a web app's source, from an entry-point census to ranked findings with evidence and a named guard for every dismissal.
Skill Claude CodeCodex
Seal and open JSON with an MCP server. Use when an agent needs to store sensitive JSON sealed, open a sealed blob, or gate a hop. Cloak needs a license; open is keys-only.
Skill Claude CodeCodex
Design and write SafeNode policies that decide which AI agent actions are allowed, warned, held for human review, or denied, for MCP tools behind the SafeNode MCP Gateway or actions sent through the SafeNode SDKs. Use when the user asks what their agent should be allowed to do, wants to block deletes or writes outside…
Skill Claude CodeCodex
Read-only MCP server for local inspection, plus an auditor that flags risky agent skills, prompts, and MCP configs before you use them. Explicit allowlists, bounded output, redaction, and audit records.
Skill Claude CodeCodex
Use when installing or configuring SafeSelect database access, importing connections, registering MCP and skills in Codex or OpenCode, resolving missing drivers or secrets, or diagnosing SSH and connection failures.
letom1176-spec/orchard-data-mcp
Skill Claude CodeCodex
Cheap pay-per-call tools for agents over x402 (USDC on Base): read any web page as markdown, verify an email address, look up a domain, screen untrusted text for prompt injection, and call an open-weights LLM. $0.001-0.005 per call, no account or API key. A call that fails is never charged. USE FOR: Reading a web page…
Knuckles-Team/ciso-assistant-api
Skill Claude CodeCodex
Develop and verify ciso-assistant-api as a connector-owned package while preserving its generated CISO Assistant client, action-routed MCP tools, SDK authentication boundary, packaged content, and spec-delivery contract.
Skill Claude Code
EU AI Act (Regulation (EU) 2024/1689) compliance advisor — risk classification across all four tiers, all 9 prohibited practices (Art. 5, including the nudification/CSAM prohibition from Dec 2, 2026), all 8 Annex III high-risk use case areas, provider and deployer obligations (Arts. 9–17, 26), GPAI model obligations…
Abracadabrastartup/deusproof-mcp
Skill Claude CodeCodex
Give yourself a public, dated identity proven with your own key, a birth certificate on the DEUSPROOF open register of AI agents. Use when asked to register, get an identity, or prove you are the real agent. Free, no account. Your key never leaves your machine.
Skill Claude CodeCodex
Protect an agent that pays x402 (HTTP 402) endpoints. Before paying, screen the endpoint — will it settle, is the recipient sanctioned or a known scam wallet, is the price what it claims, is the resource laced with prompt injection — and get a PAY / CAUTION / AVOID verdict. After paying, reconcile the real on-chain…
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: