Analyze completed agent session transcripts to produce structured retro findings and improvement proposals.
npx skills add https://github.com/kanade0404/dotfiles --skill retro-kanade0404
Please help me install this Agent Skill. Skill: retro Source: https://github.com/kanade0404/dotfiles/tree/main/.opencode/skills/retro Command: npx skills add https://github.com/kanade0404/dotfiles --skill retro-kanade0404
This Skill turns completed agent sessions into structured retro analysis so you can understand why the harness behaved the way it did and what to improve next.
Ask the retro skill to analyze the latest completed session transcript and return prioritized harness improvement proposals without making any edits.
High-intent search queries and answers about installing and using this skill.
Session transcript analysis inspects completed agent runs to identify tool usage patterns, permission denials, and loop behavior. It maps each operational finding to specific harness improvement levers like hooks, allow/deny settings, or skill edits.
Reviewing a misfired skill requires analyzing the session transcript from a fresh subagent perspective to reduce self-bias. This diagnostic approach identifies skill misfires, token waste, and blocking waits, then maps them to rule, prompt, or skill tuning recommendations.
Yes, you can generate structured retro findings and harness improvement proposals without applying edits. The analysis reviews tool statistics, permission denials, and session behavior to return prioritized recommendations for workflow quality improvements.
Diagnosing permission denials and blocking waits involves inspecting the completed session transcript and tool statistics. This operational diagnostic maps identified issues to the correct operational levers, such as allow/deny settings adjustments or new hook configurations.
A post-merge session review is needed when a PR is merged or a long-running session ends, especially if a skill misfired or wasted tokens. It produces prioritized recommendations for improving agent reliability by mapping transcript findings to harness fixes.
InnerScopeHearingCapture Claude Code session metrics and forward them to PostHog.
LSDIPPOLLCConfigure Copilot permission boundaries for agent tools and PreToolUse hooks.
ichabodcoleResearch technology integration patterns and configuration requirements into structured reports.
parcadeiNotify users of background agent progress via system reminders.
