Audit completed agent chats to identify improvements for instructions, tools, and documentation.
npx skills add https://github.com/josca42/varro --skill analyse-trajectory
Please help me install this Agent Skill. Skill: analyse-trajectory Source: https://github.com/josca42/varro/tree/main/.agents/skills/analyse-trajectory Command: npx skills add https://github.com/josca42/varro --skill analyse-trajectory
Performs a retrospective audit of completed agent chats to identify concrete improvements to instructions, tools, and documentation. This is post-hoc analysis, not an interactive exploration loop.
Run analyse-trajectory on a completed chat to generate a findings report.
High-intent search queries and answers about installing and using this skill.
Agent trajectory analysis reviews completed chats to identify actionable improvements for prompts, tool usage, and documentation, ensuring structured evaluation for system enhancements.
You can review completed agent chats to extract actionable improvements for prompts, tool usage, and documentation by applying a standardized evaluation framework to generate structured findings.
No, this performs post-hoc analysis on finished conversations and trajectories, not an interactive exploration loop. It targets completed chats to produce concise findings that inform instruction and tool changes.
The best way to prioritize fixes is applying a standardized evaluation framework to completed agent chats, which generates structured trajectory data to guide system changes and track progress over time.
No dependencies or components are required to perform the trajectory analysis. It operates independently to audit completed agent chats and generate findings reports that guide system enhancements.
lespacemanEvaluate completed WebBench tasks as PASS, PARTIAL, or FAIL using LLM-based scoring on traces and extracted data.
thoreinsteinAnalyze meeting transcripts to identify behavioral patterns and provide timestamped feedback.
DNYoussefCapture decision trajectories and extract patterns to optimize agent learning.
kivo360Classify user expertise from behavioral signals during Alfred sessions.
