The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs extensive shell command execution to manage infrastructure and run benchmarks.
Orchestrates AWS CLI commands to create VPCs, subnets, EC2 instances, and security groups in references/loader-vm-aws.md.
Interacts with the ScyllaDB Cloud REST API (api.cloud.scylladb.com) using curl to provision database clusters and manage peering in references/cluster-and-peering.md.
Uses ssh to execute orchestration scripts and Docker commands (scylladb/cassandra-stress) on remote loader VMs.
[INDIRECT_PROMPT_INJECTION]: The skill identifies a surface for indirect prompt injection by ingesting untrusted data from the user's application repository to build workload profiles.
Ingestion points: Instructions in references/workload-profile.md direct the agent to read the user's repository files (e.g., .cql files, migrations, ORM models, and request handlers) to extract schemas and queries.
Boundary markers: The ingested data is used to generate YAML configuration files for the benchmark tool; however, there are no explicit boundary markers or sanitization steps described to prevent malicious code or malformed configurations from being processed if the repository content is compromised.
Capability inventory: The skill has significant capabilities, including network access (Cloud API), infrastructure management (AWS CLI), and remote command execution on loader nodes.
Sanitization: The skill recommends a manual review via a sibling skill (scylladb-data-modeling) but does not implement automated sanitization of the extracted code before use.

scylladb-benchmark