The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of official cloud command-line interfaces (AWS CLI, Azure CLI, gcloud) and specialized security auditing tools to perform environment enumeration and security checks.
Evidence includes commands for identity verification (aws sts get-caller-identity), IAM auditing (az role assignment list), and resource scanning (gcloud compute instances list).
[EXTERNAL_DOWNLOADS]: The skill facilitates the installation and use of well-known security auditing tools from public registries.
Python packages include scoutsuite, prowler, pacu, principalmapper, roadrecon, enumerate-iam, and s3scanner via pip install.
Node.js packages include cloudsploit via npm install.
These are industry-standard tools used for cloud security posture management (CSPM) and red-teaming.
[INDIRECT_PROMPT_INJECTION]: The skill's primary workflow involves ingesting and processing data from external sources, which constitutes an attack surface for indirect prompt injection.
Ingestion points: Processes output from security scanners (Prowler, ScoutSuite, Steampipe) and cloud API responses in JSON, CSV, and HTML formats (e.g., Phase 1 in workflows/aws_full_assessment.md).
Boundary markers: No explicit delimiters or instructions to ignore embedded content within tool outputs are defined in the methodology.
Capability inventory: The skill possesses significant capabilities, including executing shell commands, writing finding reports to disk, and spawning sub-agents for specialized tasks.
Sanitization: The instructions do not define specific sanitization or filtering steps for tool outputs before they are processed by the agent's reasoning engine.

cloud-security