Back to skill

Security audit

agentic-dns

Security checks for vulnerabilities and agentic risk

Overview

The skill is for DNS operations, but it tells agents to use commands that can alter routing and intercept or block encrypted DNS traffic without clear safeguards.

Review this skill carefully before installing. It should only be used in environments where you intentionally manage DNS/firewall behavior and can tolerate or reverse network changes. Look for an implementation that shows exact firewall rules, privilege requirements, confirmation prompts, rollback/cleanup behavior, and limits on UDP/443 blocking before allowing an agent to run the mutating commands.

Vulnerability Patterns
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (1)

T05 · Unauthorized Access and Privilege Escalation

Warning
Location
SKILL.md:22
Finding
Privileged DNS Traffic Interception and Broad UDP/443 Blocking## Vulnerability Details **File Location**: `SKILL.md`, line 22 **Vulnerability Type**: `T05: Skill Instruction Hijacking` **Risk Level**: Medium ### Evidence ```text | `enforce` | `agentic-dns enforce <on\|off\|status\|clean> [target_ip]` | Intercept encrypted DNS (DoT port 853 -> pihole, DROP DoQ UDP/443). | ``` ### Technical Analysis The documented `enforce` command modifies network traffic handling by redirecting DNS-over-TLS traffic on TCP port 853 to Pi-hole and dropping UDP traffic on port 443. Firewall or packet-filter changes of this kind generally require elevated system privileges and affect traffic outside the invoking process. Dropping UDP/443 is broader than blocking only DNS-over-QUIC because UDP/443 is also commonly used by QUIC and HTTP/3. Without destination, process, interface, or protocol-aware restrictions, this rule may interrupt unrelated applications and network services. The documentation does not mention authorization checks, user confirmation, rule previews, scope limitations, backups, or automatic rollback. The package contains only `SKILL.md`; the `agentic-dns` implementation is absent. Consequently, the exact firewall commands, privilege acquisition mechanism, authentication controls, and validation behavior cannot be verified. The finding is limited to the privileged and broadly scoped behavior explicitly advertised by the Skill documentation. ### Attack Path 1. An agent loads the Skill and uses it to investigate a DNS connectivity problem. 2. The agent invokes `agentic-dns enforce on`, optionally with a target address. 3. If the external implementation runs with sufficient privileges, it installs traffic-redirection and UDP/443 filtering rules. 4. DNS-over-TLS traffic is redirected to Pi-hole, while affected UDP/443 traffic is dropped. 5. The modified rules may remain active until explicitly disabled or cleaned, disrupting encrypted DNS, QUIC, or HTTP/3 traffic within the rules' s ...[truncated 868 chars]
Remediation
## Remediation Suggestions 1. Require explicit, informed user confirmation before installing, changing, or removing firewall rules. 2. Display the exact proposed rules, affected interfaces, addresses, ports, and expected side effects before applying them. 3. Use least privilege, granting only narrowly defined firewall-management capabilities rather than unrestricted root execution. 4. Restrict blocking rules to verified DNS-over-QUIC destinations where feasible; do not use a blanket UDP/443 drop rule. 5. Scope redirection by destination address, interface, network namespace, or designated client rather than applying system-wide interception by default. 6. Validate `target_ip` as an IP address and pass it through structured firewall APIs or safely quoted argument arrays. 7. Save the prior firewall state, apply changes transactionally, and provide automatic rollback on failure, timeout, or process termination. 8. Tag created rules uniquely so that `off` and `clean` remove only rules owned by this tool. 9. Document the exact privilege requirements, rule lifetime, traffic scope, and expected effect on QUIC and HTTP/3. 10. Include the executable implementation in the reviewable package so its authorization, input validation, firewall operations, and cleanup behavior can be audited.
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (1)

Missing User Warnings

Medium
Confidence
94% confidence
Finding
The skill advertises commands that can modify live DNS routing, bypass resolvers, and enforce interception of encrypted DNS traffic, but it provides no warnings, confirmation requirements, or scope limitations. In an agent-facing skill, this is dangerous because an LLM or user may invoke disruptive network-changing actions without understanding that they can alter connectivity, privacy posture, or service availability.

Static analysis

No suspicious patterns detected.