skills/mukul975/anthropic-cybersecurity-skills/implementing-network-deception-with-honeypots/Gen Agent Trust Hub
implementing-network-deception-with-honeypots
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFECOMMAND_EXECUTIONPRIVILEGE_ESCALATIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes honeypot interaction logs which contain untrusted data provided by remote attackers such as usernames and passwords. * Ingestion points: scripts/agent.py reads log entries from /var/tmp/opencanary.log. * Boundary markers: Absent; log data is parsed and interpolated directly into the final report. * Capability inventory: scripts/agent.py can execute system commands via subprocess.run and write to the file system. * Sanitization: Absent; the script does not filter or sanitize log data before including it in JSON reports.
- [COMMAND_EXECUTION]: The management script executes system commands to control the honeypot daemon. * Evidence: scripts/agent.py calls subprocess.run to execute 'opencanaryd --start' and 'opencanaryd --status'.
- [PRIVILEGE_ESCALATION]: The deployment process involves writing to protected system directories and recommends the use of elevated permissions for installation. * Evidence: scripts/agent.py writes to /etc/opencanaryd/opencanary.conf and references/api-reference.md includes sudo commands.
- [EXTERNAL_DOWNLOADS]: The skill documentation recommends downloading and installing third-party software and container images. * Evidence: references/api-reference.md provides instructions for pip install and docker pull for the OpenCanary tool.
Audit Metadata