The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads official CLI tools and installers from well-known official sources, including Amazon's AWS CLI and Google's Cloud SDK.
[REMOTE_CODE_EXECUTION]: Executes an installation script directly from Google's official SDK endpoint (sdk.cloud.google.com) via a pipe to bash. As this originates from a well-known service provider, it is considered a legitimate installation step.
[COMMAND_EXECUTION]: Utilizes a wide range of administrative CLI tools, including AWS CLI, gcloud, and Azure PowerShell modules, to interact with and audit cloud infrastructure.
[DATA_EXFILTRATION]: Provides procedures for extracting sensitive data such as IAM credentials, environment variables from serverless functions, and secrets stored in Azure Key Vault as part of its security assessment functionality.
[PRIVILEGE_ESCALATION]: Details methods for escalating access within cloud environments, such as manipulating service principal roles and permissions, which is consistent with the skill's stated purpose of penetration testing.
[PERSISTENCE]: Includes instructions for establishing persistence through the creation of secondary administrative users and access keys within the targeted cloud environments.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: Processes data retrieved from external cloud APIs, metadata services, and security auditing tool outputs like ScoutSuite results in SKILL.md and references/advanced-cloud-scripts.md.
Boundary markers: Does not utilize explicit delimiters to distinguish between benign resource data and potentially malicious embedded instructions within cloud metadata.
Capability inventory: The skill possesses extensive capabilities across multiple cloud platforms, including network operations, file system access, and high-privilege API interaction.
Sanitization: Lacks explicit validation or escaping for data ingested from external cloud environments before it is processed by the agent.

cloud-penetration-testing