The Agent Skills Directory

[COMMAND_EXECUTION]: The documentation instructs the agent to run a local script, skills/skill-eval/scripts/eval_runner.py, to validate and execute skill benchmarks. This is a standard and documented developer workflow for quality measurement.
[PROMPT_INJECTION]: The skill is designed to process external feedback and session history, which creates a surface for indirect prompt injection. However, it explicitly includes robust safety protocols in references/diagnostic-and-safety-findings.md to identify and isolate malicious instructions by deriving safety predicates.
[SAFE]: The skill promotes security best practices, such as credential redaction and prevention of sensitive data flows at the earliest boundary, and does not contain any unauthorized network activity or remote code downloads.

skill-quality