The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs extensive infrastructure operations using the AWS CLI and kubectl to manage chaos experiments.
Evidence: Uses aws fis start-experiment, aws fis stop-experiment, and aws cloudformation delete-stack to modify cloud resources.
Evidence: Includes shell script logic for status polling in references/cli-commands.md which executes multiple commands in a loop.
[DATA_EXFILTRATION]: The skill accesses and aggregates sensitive system and application data for reporting.
Evidence: Utilizes kubectl logs -f to collect real-time application logs and AWS CLI to fetch managed service logs.
Evidence: Reads environment-specific configuration and CloudFormation outputs to resolve experiment targets.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because it processes untrusted external data.
Ingestion points: Reads experiment metadata from directory-level README.md files and analyzes application log streams.
Boundary markers: No explicit delimiters or instructions are used to prevent the agent from obeying instructions potentially embedded in logs or config files.
Capability inventory: The agent has permissions to delete stacks, stop experiments, and execute shell commands via referenced scripts.
Sanitization: There is no evidence of log sanitization or validation of the README content before it is used to drive the workflow.

aws-fis-experiment-execute