skills/mukul975/anthropic-cybersecurity-skills/conducting-cloud-penetration-testing/Gen Agent Trust Hub
conducting-cloud-penetration-testing
Warn
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONPERSISTENCEPRIVILEGE_ESCALATIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The scripts/agent.py script leverages subprocess.run to execute multiple AWS CLI commands for resource enumeration.
- [PERSISTENCE]: The SKILL.md file contains explicit instructions for creating persistence artifacts, such as a 'backdoor IAM user' with AdministratorAccess and associated access keys as part of persistence and detection evasion testing.
- [PRIVILEGE_ESCALATION]: The skill provides methodologies for identifying and exploiting IAM misconfigurations, including script logic to check for iam:CreatePolicyVersion permissions which allow for escalation to administrative privileges.
- [DATA_EXFILTRATION]: The scripts/agent.py script identifies and collects potentially sensitive information, including IAM role trust relationships and secrets stored in Lambda environment variables by scanning for keys such as password, secret, and api_key.
- [INDIRECT_PROMPT_INJECTION]: (1) Ingestion points: External data is ingested in scripts/agent.py from the outputs of AWS CLI commands. (2) Boundary markers: No delimiters or instructions to ignore embedded content are used. (3) Capability inventory: The skill possesses command execution capabilities (subprocess.run) and file-writing capabilities. (4) Sanitization: The script parses command output into JSON but does not sanitize or filter the content of the fields before inclusion in the final report.
Audit Metadata