Audit Libra Legal AI's multi-agent system across pipeline stages and generate actionable findings.
npx skills add https://github.com/ossarg/ali --skill system-audit-ossarg
Please help me install this Agent Skill. Skill: system-audit Source: https://github.com/ossarg/ali/tree/main/skills/system-audit Command: npx skills add https://github.com/ossarg/ali --skill system-audit-ossarg
Audit Libra Legal AI's multi-agent system to identify deviations, validate outputs, and surface actionable improvements across cases and pipeline stages.
Trigger a post-pipeline audit immediately after a completed case and review the results in daily logs.
High-intent search queries and answers about installing and using this skill.
To audit a multi-agent system, run a structured evaluation that scores per-sub-agent performance, analyzes pipeline handoffs, and surfaces actionable improvements aligned with your orchestration guidelines.
A post-pipeline case audit is a structured review triggered after a completed case to validate outputs, identify deviations, and generate actionable findings for continuous improvement and compliance.
Perform a system health review by applying a periodic evaluation framework that scores sub-agents, updates regression artifacts, and generates recommendations to ensure continuous compliance.
Yes, you can score individual sub-agents during an orchestration audit by applying structured multi-step evaluation criteria that validate outputs and surface actionable improvements per pipeline.
The best way to surface actionable improvements from autoevolution is to conduct deep-dive audits that evaluate deviations, update calibration logs, and generate comprehensive compliance reports.
Without structured evaluation dimensions, auditing multi-agent systems lacks per-sub-agent scoring and handoff analysis, preventing the generation of actionable findings and regression updates needed for continuous improvement.
yu13130122297Identify uncertainties in research framing, datasets, metrics, and baselines.
augustoheissAudit machine learning models for reproducibility, calibration, performance, and fairness.
GodSpeedAICoordinate PM-Agent workflows for case management, task delegation, and artifact pipelines.
dalawwaCoordinate multi-agent generation, multi-judge evaluation, and evidence-based synthesis for task solutions.
