The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a set of vetted helper scripts to perform resource discovery and assessment. These scripts execute sanctioned local tools such as the aws CLI and terraform. All AWS operations are restricted to a read-only allowlist (e.g., get-, list-, describe-), and the agent is explicitly forbidden from composing custom shell pipelines or using untrusted interpreters.\n- [DATA_EXFILTRATION]: Network activity is restricted to legitimate AWS service endpoints through the official CLI. All workload data, metadata, and generated reports are stored locally on the user's filesystem and are not transmitted to any non-whitelisted or external domains.\n- [SAFE]: The skill includes internal security audit scripts and enforces a strict path convention for script execution. The author (aws-samples) and all referenced documentation (e.g., NIST, FinOps Foundation, AWS) are trusted or well-known entities. No evidence of obfuscation, credential harvesting, or prompt injection was found.

aio2-review