github.com/digitalfortressaudit
Security tools for teams that ship AI agents
Practical AI-agent, MCP and Kubernetes security tools and audit packs. Start with free read-only checkers like mcp-tool-pin-check and agent-k8s-harden-check, then go deeper with prompt and template packs you can use the same day.
Free tools
Open source on GitHub under digitalfortressaudit.
-
mcp-tool-pin-check
Read-only checker that pins MCP tool definitions and diffs later listings for drift.
-
agent-k8s-harden-check
Read-only checker for Kubernetes AI-agent workloads. Flags a missing gVisor runtime class, service-account token automount, privileged or root containers, host namespaces, missing limits, and missing metadata or RFC1918 egress NetworkPolicy.
-
agent-guardrails-baseline
Free 15-minute AI agent guardrails baseline: 25 checks across identity, egress, tool authorization, sandbox, and logging.
-
ai-security-starter-checklist
Free AI/LLM security starter: 5 copy-paste audit prompts + a read-only Python repo scanner.
-
ai-system-agent-inventory-register
Free AI system & agent inventory register (CSV + Excel). NIST AI RMF, ISO/IEC 42001 & EU AI Act screening fields—including MCP tools and delegated permissions.
Paid packs
Audit prompts, templates and worksheets, sold on Gumroad.
-
AI Agent Governance & Adverse-Decision Compliance Kit $799
Editable governance kit: risk classification, impact assessments, agent action boundaries, incident response, and EU AI Act Art. 50 / Colorado / California adverse-decision notices.
-
Agent Incident Response Pack $349
Detect, halt, escalate and evidence agent incidents: scored severity matrix, stop-when-blocked rules, outbound limits and an agent logging schema.
-
Multi-Agent Orchestration Blueprint $149
System-prompt architectures, handoff state machines and a Notion operations workspace for running a team of agents. Runtime-agnostic.
-
AI Security & Infrastructure Prompt Pack $99
Copy-paste audit prompts, multi-agent threat-model templates and read-only scripts for LLM agents, MCP servers, Kubernetes and cloud.
-
Agent Runtime Containment Pack $99
Review the runtime boundary of a tool-using agent: egress, identity, Kubernetes sandboxing and out-of-model tool-call authorization on AWS or Kubernetes.
-
Coding Agent Repo Trust Pack $99
Decide whether a coding agent may open a repo and load its rule files, MCP config, hooks, skills and packages. 30 prompts, 4 templates.
-
MCP & Multi-Agent Authorization Security Pack $79
Decide which MCP servers to trust, how far an agent may delegate, and what evidence shows a delegated action was authorized.
Contact
Questions about a tool or pack: digital.fortress.audit@gmail.com