Adversarially test an LLM agent before it ships: prompt-injection and exfiltration lenses, latency and token-budget lenses, and a scored regression suite you can re-run.
Already have an account?
Install this skill from your terminal in one command.
npx promptdock@latest install creator2/agent-test-harnessA repeatable adversarial suite for an agent that has tools. Two security lenses and two performance lenses, each with concrete probes, scored into a regression baseline you can diff on the next release.
Before giving an agent a tool that writes, spends, or sen
Sign in to read the full skill
This is a premium skill. Create a free account to see the full SKILL.md; installing it needs an Explorer plan or higher.