Analyze Swarm trajectories, tool usage, and orchestration to detect inefficiencies.
npx skills add https://github.com/dewitt/swarm --skill quality-review-dewitt
Please help me install this Agent Skill. Skill: quality_review Source: https://github.com/dewitt/swarm/tree/main/skills/quality_review Command: npx skills add https://github.com/dewitt/swarm --skill quality-review-dewitt
The Quality Reviewer provides a rigorous, end-to-end evaluation agent focused on analyzing Swarm's cognitive trajectories, tool-use discipline, handoff quality, and adherence to grading rubrics. It helps identify loops, inefficiencies, and orchestration gaps that hinder task success.
Initiate a comprehensive quality review of the active Swarm session and produce actionable improvement recommendations.
High-intent search queries and answers about installing and using this skill.
Evaluating agent orchestration involves analyzing Swarm trajectories and telemetry to detect loops, inefficiencies, and sub-optimal tool calls. This review identifies misrouted handoffs and orchestration gaps that hinder task success.
Trajectory analysis in multi-agent systems examines how agents plan, execute, and monitor tasks to flag cognitive dead ends. It improves workflow quality by identifying inefficient tool usage, brittle prompts, and orchestration gaps, generating actionable remediation guidance.
Auditing SKILL.md instructions and sub-agent routing involves reviewing how sub-agents are selected and evaluating their instructions for clarity and bounded behavior. This skill audit detects brittle prompts and ensures proper persona adherence within the Swarm session.
No external dependencies are required to collect local telemetry. The quality review tool leverages local file access, directory listing, text search, and shell execution via built-in functions like list_local_files, read_local_file, grep_search, and bash_execute to gather data.
Yes, you can use this to inspect evaluation rubrics and test fixtures. The tool evaluates available rubrics, fixtures, and scenarios to ensure objective, measurable assessments and groups findings into actionable categories with suggested experiments.
hiddink-aiConfigure OpenTelemetry environment variables for Claude Code usage metrics and event logging.
SinShine5Generate, write, and run goal-based evaluation tests for agents using MCP tools.
chiptoe-svgAdd a local monitoring dashboard for NanoClaw deployments.
flutterOrchestrate automated evaluation of agent skills using isolated subagents and rubrics.
