Trace user corrections to root causes and audit subsystem files for fixes.
npx skills add https://github.com/nikolaj-lat/World-Puppeteer --skill reflect-nikolaj-lat
Please help me install this Agent Skill. Skill: reflect Source: https://github.com/nikolaj-lat/World-Puppeteer/tree/main/.claude/skills/reflect Command: npx skills add https://github.com/nikolaj-lat/World-Puppeteer --skill reflect-nikolaj-lat
Reflect helps ensure AI behavior remains robust by automatically detecting and addressing corrections across the system, reducing repetitive mistakes and governance gaps.
Describe the correction you want fixed, then trigger Reflect to audit the current subsystem and propose changes.
High-intent search queries and answers about installing and using this skill.
To audit AI agent rules and fix repetitive mistakes, you identify user corrections to extract root causes, then audit subsystem files like skill definitions and agent configurations to propose targeted, approved changes resolving underlying gaps.
Correction propagation works by tracing user feedback to root causes and assigning durable fixes. It audits related subsystem files to uncover governance gaps linked to the correction and proposes minimal changes for future reliability.
The best way to trace user feedback to root causes in agent configurations is to extract intended outcomes from corrections, then comprehensively audit skills, references, rules, scripts, and CLAUDE.md files to map latent issues linked to the feedback.
Yes, you can propose auditable edits to CLAUDE.md and agent rules through change-driven governance. The system proposes minimal, targeted edits to skills, agents, or rules with built-in approvals to safely fix root causes.
Yes, this correction subsystem works with existing skill definitions and scripts by comprehensively auditing all related subsystem files, including references, rules, and CLAUDE configurations, to surface latent issues and propose minimal approved changes.
meainReview conversations to derive lessons and propose CLAUDE.md guideline updates.
yachelaClassifies proposed AI use cases as approved, conditional, or not approved against a governance registry and red-line rules.
chaimtConvert ADRs into a DAG-based Architecture Description with a three-phase workflow.
Ilenburg1993Audit code for semantic flaws and generate a Markdown report with corrections.
