analysis-tshark

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly instructs extracting and decoding credentials from packet captures (HTTP Basic auth, FTP passwords, NTLM hashes, etc.) and shows commands that output those secrets (e.g., piping http.authorization to base64 -d), which requires the agent to handle and potentially emit secret values verbatim.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.85). The skill’s runtime workflow ingests packet-capture content (e.g., via tshark -r capture.pcap ... and --export-objects ...), and those captures can contain arbitrary outsider-authored free text from network traffic (HTTP/DNS/SMB payloads) that becomes readable prose in the LLM context.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The CI template in the skill contains a runtime remote-install command that fetches and executes code (curl -s https://raw.githubusercontent.com/aquasecurity/tfsec/master/scripts/install_linux.sh | bash) and also references GitHub Actions (e.g., aquasecurity/trivy-action@master, actions/checkout@v4) which pull and run external code during CI execution, so external content would be executed at runtime.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 0.90). This skill explicitly instructs using sudo for packet capture and includes commands that modify system state and privileges (e.g., sudo tshark, sudo usermod -a -G wireshark $USER, sudo setcap ...), which pushes the agent to obtain elevated privileges and alter the host configuration.