v0.6.0 — 2026-02-25

Feedback: Report Wrong AI Responses from Sessions

You can now report bad AI responses directly from a skill's detail page, using real chat history as evidence.

• Report issue button — Every skill now has a "Report issue" action that opens a dedicated feedback workspace.
• Session list — The workspace shows all Cursor and Codex sessions where the selected skill was used, so you can find the conversation with the bad response.
• Dialogue view — Opening a session shows the full user/AI conversation. Messages before the first skill invocation are collapsed by default to keep focus on what the agent actually did; a toggle reveals earlier context.
• Skill invocation badges — Skill invocations appear as labelled badges rather than raw SKILL.md file paths.
• Mark wrong response — Select any AI message to mark it as wrong. Then describe what was wrong and what the correct behaviour should have been. An optional suggested rule can be included.
• AI analysis — After filling in the expected behaviour, the app runs an AI analysis of the issue: likely cause, whether a rule fix is feasible, any contradictions with existing guidance, and a suggested rule patch.
• Local draft + GitHub issue — Reports are saved locally first as "Pending sync". A separate "Submit report" action creates a GitHub issue in the skill's repository with the full report attached.