The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to install several third-party security tools from repositories not included in the internal trusted list, specifically Pacu (from Rhino Security Labs via pip), CloudFox (from Bishop Fox via go install), and Principal Mapper (from NCC Group via pip).
[COMMAND_EXECUTION]: The assessment workflow requires the execution of multiple CLI tools and shell commands, including the AWS CLI for environment enumeration, Pacu for automated exploitation scanning, and CloudFox for attack path identification. It also includes a custom Python one-liner in the workflow to parse trust policies.
[INDIRECT_PROMPT_INJECTION]: The skill and its included script (agent.py) process AWS IAM policies and role trust documents which are retrieved from the target environment and could theoretically be controlled by an external actor.
Ingestion points: IAM user policies, role trust policies, and group memberships retrieved via the iam.get_policy_version and iam.list_attached_user_policies methods in scripts/agent.py.
Boundary markers: None identified; the skill assumes the integrity of the data returned by the AWS API.
Capability inventory: The skill possesses the capability to perform AWS API calls via boto3 and execute the shell commands described in the workflow.
Sanitization: No explicit validation or sanitization of the JSON-formatted IAM policies is performed before checking for dangerous permission strings.

performing-aws-privilege-escalation-assessment