Automate AIDD evaluation harness execution and generate audit reports.
npx skills add https://github.com/GrinRus/ai_driven_dev_v2 --skill aidd-eval
Please help me install this Agent Skill. Skill: aidd-eval Source: https://github.com/GrinRus/ai_driven_dev_v2/tree/main/.agents/skills/aidd-eval Command: npx skills add https://github.com/GrinRus/ai_driven_dev_v2 --skill aidd-eval
Manually running evaluation scenarios for ai_driven_dev_v2 (AIDD) is time-consuming, error-prone, and often produces inconsistent, non-inspectable audit records. This Skill eliminates that overhead by providing a standardized, governed workflow for executing eval harnesses, validating stage outputs against document contracts, and producing durable, reproducible audit artifacts for both deterministic and manual eval lanes.
Use the aidd-eval skill to run the sqlite-utils live E2E scenario against the OpenCode runtime and generate a full audit report with validator outcomes and failure analysis.
High-intent search queries and answers about installing and using this skill.
Automate AIDD evaluation harness execution by running standardized workflows that validate stage outputs against document contracts. This eliminates manual test orchestration and ensures consistent artifact collection across both deterministic fixture-seed and manual live authored-task-pool eval lanes.
Validate document contracts by automatically checking all AIDD stage output Markdown files against predefined rules. This process catches formatting, content, and structural errors before they propagate, ensuring stage outputs strictly comply with the framework's specified contracts.
Runtime log capture works by preserving raw logs, normalized events, validator reports, and repair history during eval runs. This comprehensive evidence capture ensures full auditability of self-repair behaviors by separating infrastructure failures from model or document contract violations.
Yes, you can use it for both deterministic fixture-seed scenarios in CI environments and manual live authored-task-pool scenarios for real-world runtime testing. It supports all AIDD runtimes, providing standardized execution verdicts and failure analysis across both eval lane types.
Durable audit artifacts are needed to separate infrastructure failures from model or document contract violations during analysis. They provide reproducible, inspectable evidence including runtime logs and validator outcomes, replacing inconsistent manual data collection with a standardized execution verdict taxonomy.
humanity4aiProvide a test fixture with too few scenarios for eval harness validation.
ramtinJ95Diagnose dbt Cloud job failures using MCP tools, logs, and artifacts.
GrinRusDefine and enforce shared output, read, and loop-safety policies across AIDD stages.
EnesMeyzin98Transform experiment metrics into advisory governance classifications with supporting evidence.
