The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/record.sh script executes shell commands provided as arguments via sh -c to facilitate the recording of 'red' (failing) and 'green' (passing) test runs. This is the core functionality of the test-recording tool.
[DYNAMIC_EXECUTION]: The scripts/record.sh script allows for the execution of dynamic 'predicate' commands to verify the state of the environment during test cycles.
[INDIRECT_PROMPT_INJECTION]: The skill scripts process local files and command outputs, creating an ingestion surface for potentially untrusted data.
Ingestion points: The scripts read the measurements.jsonl ledger and capture the standard output/error of executed test commands in scripts/record.sh.
Boundary markers: Data is structured using JSON and JSONL formats.
Capability inventory: The skill utilizes shell execution, file system operations, and network requests via curl to interact with the vendor's simulation environment.
Sanitization: The scripts/ledger.sh utility includes logic to scan for and block the inclusion of sensitive credential patterns (e.g., AWS keys, Slack tokens) in the ledger, serving as a security safeguard.
[SAFE]: All network operations are directed towards the vendor's official API domain (svc.api.veris.ai). The behavior of the scripts and documentation is entirely consistent with the primary purpose of providing a developer reference and verification toolkit.

veris-reference