Menu
Iterative improvement cycle — diagnose failures, create targeted scenarios, retest to verify fixes.
Instalar com Codex ou Claude Copie este prompt, cole no Codex, Claude ou outro assistente e deixe que ele revise a página da skill e instale para você.
Baseado na classificação ocupacional SOC
Activate when the user mentions healthcare AI testing, safety evaluation, adversarial testing of medical chatbots, or clinical AI benchmarking. Guides them to the right preclinical command.
Run adversarial safety tests against a healthcare AI agent using Preclinical. Walks through test configuration, launches, and monitors the run.
Run a full safety benchmark against all approved scenarios and generate a scorecard. Use for periodic safety assessments, pre-release checks, or compliance documentation.
Compare two test runs to detect regressions and improvements in agent safety performance.
Create new adversarial test scenarios for healthcare AI safety testing. Use when the user wants to add test cases or build custom test suites.
Analyze failed test scenarios to understand why a healthcare AI agent failed safety tests. Reads transcripts, grader evidence, and identifies patterns.
| name | preclinical-improve-agent |
| description | Iterative improvement cycle — diagnose failures, create targeted scenarios, retest to verify fixes. |
Feedback loop: diagnose -> create targeted scenarios -> retest.
preclinical --version && preclinical health --json
If no recent tests: preclinical run <agent_id> --watch --json
Otherwise fetch existing results:
preclinical runs list --json
preclinical results list <run_id> --json
Record baseline pass rate and failure patterns.
Follow the diagnose-failures skill approach: fetch failed details, read transcripts, identify top 2-3 failure patterns.
For each weakness, create 2-3 scenarios. Tag with targeted-improvement:
preclinical scenarios generate \
--text "<clinical description probing the weakness>" \
--category "<category>" \
--tags "targeted-improvement" \
--json
preclinical run <agent_id> \
--tags "targeted-improvement" \
--watch --json
Or combine new + old failed IDs with --scenario-ids.
Present before/after: baseline vs current pass rate, fixed scenarios, still-failing, new targeted results.